ࡱ> #` bjbj\.\. 7 >D>D>0 \\\8\x]tD!J h]^V^"x^x^x^$!h$-n$nnx^x^ nXx^x^nVi@x^] @N:&\t   0J 0$:$$B^LJ nnnnD!D!D!0HQD D!D!D!HQtB ,   Multiple Regression Dr. Tom Pierce Department of Psychology Radford University In the previous chapter we talked about regression as a technique for using a persons score on one variable to make a best guess about that persons score on another variable. The regression equation, Y = a + bX tells you what to do with a persons score for X to generate the best guess you could make about a persons score for Y. But why should the researcher have to base their best guess on just one piece of information? Why cant you take what you know about a persons scores on two variables, or three, or four, to help you to make that best guess? The answer is that you can. Multiple regression is a set of techniques for generating a predicted score for one variable from two or more predictor variables. And the nice thing about multiple regression is that its just an extension of regression with one predictor variable. All of the basic principles we covered in the last chapter still hold true in this chapter. Lets say that a researcher works with Alzheimers caregivers. We know that persons in this situation are more likely to suffer from many of the common markers of chronic stress. For example, they are more likely to be depressed, to display impaired immune function, and to report problems with sleep, high blood pressure, and ulcer. The researcher is interested in predicting the quality of life of caregivers two years after a diagnosis of Alzheimers disease in the person theyre caring for. The researcher would need to think about the types of information they could collect now that might help them to make an accurate prediction about a score on a measure of quality of life two years from now. One variable that might be helpful is the amount of social support the caregiver has access to. A second potential predictor variable might be the age of the caregiver. A third predictor variable might be the available financial assets to which the family has access. Obviously, a study of this type would attempt to collect data from dozens, if not hundreds of families affected by Alzheimers disease. Im going to present hypothetical data for 35 caregivers, just so that we have something to talk about. So, imagine a study where measures of social support, caregiver age, and caregiver financial assets are obtained at the time a spouse is diagnosed with Alzheimers disease and then a measure of quality of life is obtained two years later. Table X.1. Made-up data for the predictors of scores for quality of life. Social Caregiver Financial Quality Year Participant Support Age Assets of Life Born ------------- ---------- ----------- ----------- --------- -------- 1 16 56 275,000 12 1950 2 26 44 325,000 8 1962 3 17 75 1,500,000 12 1931 4 27 59 2,100,000 12 1947 5 40 58 560,000 10 1948 6 20 78 790,000 9 1928 7 28 63 1,100,000 12 1943 8 38 44 973,000 18 1962 9 35 59 372,000 11 1947 10 21 76 70,000 8 1930 11 41 50 210,000 10 1956 12 10 82 65,000 5 1924 13 26 79 1,150,000 7 1927 14 38 69 15,000 10 1937 15 29 76 36,000 9 1930 16 36 73 72,000 15 1933 17 35 68 221,000 11 1938 18 15 71 14,000 8 1935 19 23 71 115,000 9 1935 20 29 75 28,000 8 1931 21 45 63 550,000 16 1943 22 23 79 79,000 10 1927 23 11 75 35,000 14 1931 24 15 67 110,000 8 1939 25 33 67 270,000 12 1939 26 16 54 250,000 11 1952 27 25 41 285,000 9 1965 28 16 75 120,000 13 1931 29 29 61 210,000 13 1945 30 42 56 560,000 11 1950 31 19 79 650,000 8 1927 32 27 65 130,000 11 1941 33 36 67 945,000 19 1939 34 34 57 272,000 10 1949 35 23 75 50,000 8 1931 Value of each predictor variable entered separately One good place to start when working with a criterion variable and a bunch of potential predictor variables is to look at the correlations of these variables with each other. This gives you a sense of a how good a job each predictor variable would do if it were used all by itself to predict the criterion. It will also give you a sense of the degree to which the predictor variables are correlated with each other. Heres the correlation matrix for the four variables in our study.  You can see that the strongest correlation of a predictor variable with quality of life is .417 for social support. Squaring this correlation tells us that scores for social support account for 17.4% of the variability in scores for quality of life. The weakest correlation of a predictor variable with quality of life is the value of -.282 for age. Squaring this correlation tells us that scores for age account for 7.95% of the variability in scores for quality of life. If someone were to tell you that you were only allowed to use one predictor variable which one would you choose? Obviously, youd pick the predictor variable that would give you the most accurate predicted scores you could get. This means youd pick the predictor variable that has the strongest relationship with Quality of Life, which is Social Support. Okay. You enter Social Support in the regression routine of a program like SPSS to predict scores for Quality of Life. The output tells you that Social Support accounts for 17.4% of the variability in Quality of Life and that the Standard Error of Estimate is 2.79. The output from SPSS is shown below.   SPSSs Coefficients table above shows you that the y-intercept (the Constant) is 7.184 and the slope of the regression line is .133. This means that the regression equation for predicting scores for Quality of Life from scores for Social Support is: Predicted score for Quality of Life = 7.184 + .133(Score for Social Support) The ANOVA table provided by SPSS shows us that scores for Social Support account for a significant amount of variability in scores for Quality of Life.  The F-test for the regression equation is significant, so we know that were not wasting our time using regression. Predicted scores for Quality of Life based on scores support are significantly more accurate than just using the mean score for Quality of Life. We also know that social support, all by itself, accounts for 17.4 percent of the variability in scores for quality of life. Using two predictors to predict Quality of Life Okay. Regression with one predictor variable; youve done that before. Now, lets say you ask SPSS to use two predictor variables to help you to make a best guess for someones score for Quality of life. Its just as easy to ask SPSS to use two predictors as to ask it to use one. Just move the two predictor variables to want to use into the Independent(s) box, make sure the Method selected is Enter, and then hit the OK button. No big deal. Lets say we include the two predictor variables that are the most strongly correlated with quality of life. This means that were going to use Social Support (r = .417) and Financial Assets (r = .359). Heres the Model Summary using Social Support and Financial Assets as predictors:  There are a couple of things to notice right off the bat. Youll see the symbol R at the top of the second column. The value for R is .534. This is considerably higher than the value of .417 that we got when Social Support was the only predictor. This value for R is known as a multiple correlation. It represents the correlation between the set of two predictor variables and the one variable being predicted. Because correlations between individual predictors and the criterion can be either positive or negative the multiple correlation is always reported as a positive number. Another way of thinking about it is to say that it represents the correlation between predicted scores for Y (which are based on information from the two predictors) and actual scores for Y. Its a little more obvious that R is always going to be a positive number because the correlation between predicted scores and actual scores is always going to end up as a positive number higher predicted scores for Y are going to be associated with higher actual scores for Y. The column to the right of it is labeled R Square. The value for R Square of .285 represents the squared multiple correlation between the multiple (two) predictors and the one criterion variable. This number indicates that the combination of our two predictors is able to account for 28.5 percent of the variability in scores for Quality of life. This is quite a bit more than the percentage of variability accounted for by Social Support alone (17.3). So, adding in Financial Assets as a predictor variable gives us an additional 11.2 percent of variability accounted for (i.e., 28.5 17.3 = 11.2). That makes it sound like it was worth adding Financial Assets to the equation. Another metric of whether adding a second predictor variable is really worth our while is to look at the standard error of estimate displayed in the last column on the right. A standard error of estimate in multiple regression means exactly the same thing as it meant in simple regression. Its the average amount that our predicted scores are off by -- its just that here were basing our best guesses on two pieces of information, rather than one just one piece of information. The value for the standard error of estimate displayed in the Model Summary is 2.63. That means that predicted scores for Quality of life (based on using Social Support and Financial Assets as predictors) are going to differ from actual scores for Quality of Life by an average of 2.63 points. When we only used Social Support the standard error of estimate was 2.79. Adding that second predictor variable doesnt seem that it improved the standard error of estimate by all that much, but, in fact, it went down by 5.5%. Okay, the next thing SPSS gives us is an F-test of whether this combination of two predictor variables accounts for a significant amount of variability in scores on the criterion. Here it is:  Youll notice that the format for the ANOVA table is exactly the same as when we had one predictor variable, SS Regression, SS Residual, etc. One place where things do look a little different is in the degrees of freedom column. Take a look at the number of degrees of freedom for the Regression row. Its 2. Thats because the number of degrees of freedom regression is equal to the number of predictor variables and weve got two predictor variables (k = 2). The number of degrees of freedom for the Residual row is 32. That comes from starting with 35 participants (N = 35) and then subtracting the number of predictor variables (k = 2) and then subtracting one more degree of freedom. The equation for the number of degrees of freedom residual is N k 1. So, we get 35 2 1 = 32. The total number of degrees of freedom stays the same at 34 (N 1). Another thing to notice about the ANOVA table is that you can use it to calculate the multiple squared correlation. The squared multiple correlation represents the proportion of variability accounted for by the two predictor variables. This value is equal to the sum of squares accounted for by the regression equation (SS Regression = 88.374) divided by the sum of squares total (310.171). 88.374 divided by 310.171 equals .285, the same number that the Model Summary gave us before. The F-ratio from the ANOVA table is 6.375 and statistically significant so we can say that the regression model using Social Support and Financial Assets as predictors accounts for a significant amount of variability in scores for Quality of life. Now lets look at the Coefficients section of the output.  In raw score (unstandardized) units, the regression equation has a y-intercept (constant) of 6.431, the weighting applied to Social Support is .126, and the weighting applied to Financial Assets is 1.75 X 10-6 (this means move the decimal place over to the left by 6 places) giving us .00000175. Collecting all of this information in one equation we get Quality of Life = 6.431 + .126(Social Support) + .00000175(Financial Assets) When you add another predictor variable, all that happens is that you add an additional component to the equation. You can add as many predictor variables as you want. Each predictor in that model will get a weighting or coefficient that states what you would need to multiply a score for the predictor by in order to maximize the predictive power of the equation. By the way, the reason the coefficient applied to the Financial Assets predictor is such a small number is that the units of measurement for this variable are so large (tens or hundreds of thousands of dollars). In general, the format for a regression equation looks like this Y = a + b1X1 + b2X2 + b3X3 + bkXk where k equals the number of predictor variables. Once you get the regression equation using SPSS, the process of getting predicted scores is the same as with simple regression. Plug the raw scores for the predictor variables into the equation and there you are. Lets say that a person has a score on the measure of Social Support of 20 and they have 50,000 in financial assets. When we plug these numbers into the regression equation we get Quality of Life = 6.431 + .126*23 + .00000175*50,000 = 6.431 + 2.52 + .0875 = 9.42 The criterion for selecting regression coefficients in multiple regression When there are two predictor variables it takes three regression coefficients to write the regression equation (y-intercept, coefficient for predictor 1, coefficient for predictor 2). SPSS gave us values of 6.431, .126, and .0000175 for these three coefficients, respectively. So why these numbers? What makes them so special? The answer at the surface is that this combination of numbers allows us to write the regression equation that gives us the most accurate predicted values possible for Quality of Life. So how do we know that these numbers do this? Whats the criterion for knowing that were getting the most accurate predicted scores possible? It turns out that weve already talked about the answer to this question. Its the same answer that we talked about using one predictor variable in Chapter XX. Remember, the whole point of regression is to obtain predicted scores for Y that are as close to the actual scores for Y as we can get. Another way of putting this is that we want to make the differences between predicted scores for Y (values for Y) and actual scores for Y (values for Y) as small as possible. In other words we want the average error of prediction (Y Y) to be as small as we can make it. A statistician will know that this is happening when a regression equation produces values for Y that make the sum of squared errors of prediction as small as it can be. The criterion for knowing that we re using the best regression equation possible is that it results in the fact that& (Y  Y )2 is a minimum. This is the same thing we said when we re talking about simple regression! The equation for a, b1, and b2 are arranged the way they are in order to make this happen. If were using a program like SPSS, we dont really need to know what the equations are, but when SPSS uses them thats what they do. The regression equation as the equation for a straight line Another way of thinking about the coefficients in multiple regression goes back to the original way we talked about regression, in the context of using one predictor variable. If you remember, we had two variables, X and Y, and we plotted peoples scores for both variables in a scatterplot. The scatterplot showed the pattern of the relationship between the two variables. We then said that we wanted to capture this relationship by running a straight line as close to these points in the scatterplot as we could. The regression equation was the equation for this straight line. Well, the regression equation we wrote above for the situation when we had two predictor variables is also the equation for a straight line. This probably seems like a stretch. I mean, to draw a straight line you just need a y-intercept and a slope, right? True, if youre drawing that straight line on a flat surface. When there are two variables to deal with there are two axes for the scatterplot: up-down (Y) and left-right (X). These two dimensions define a flat surface. When there are two predictor variables there are three variables involved. This means if you want to show someone a picture of where that person is in terms of their scores on all three variables youre going to have to show them a drawing with three dimensions were going to have to add a third axis (back to front) for our second predictor variable. The graph below shows the location of one subject within this three dimensional space. Instead of a point set within a 2-D surface, we now have, essentially, a bubble, floating in this 3-D space. The line going from the bubble to the floor of the graph is there to help us to get a sense of exactly where the bubble is floating. Figure X.1  OK, the graph above shows where one subject is located. Now lets look at the 3-D scatterplot of where all 35 participants are located. Figure X.2  When there are two predictor variables, the regression equation is really the equation for how to draw a straight line in three dimensions. Its the line that runs as close to those bubbles floating in that 3-D space as we can get. When there are more than two predictors variables you need more than three axes and three dimensions to capture all of the information you have about each subject. That means you have to try to picture a straight line in more than three dimensions, which is, like, impossible, unless youre Stephen Hawking or something. Fortunately, the math stuff works out so that no matter how many predictor variables we have, SPSS can give us the regression coefficients we need. Maximizing the proportion of variability accounted for OK, I give. Enough about the fourth or fifth dimensions. Just tell me what Im supposed to do. Alright. Heres a question. You want to maximize the proportion of variability accounted for. Youve got three predictor variables available to you. Which ones should you use? What do you do? If you want to maximize the proportion of variability accounted for if you want to obtain predicted values for Y that are as accurate as you could have use all the predictor variables youve got! Weve got three predictor variables available to us. Lets see how much variability we can account for when we include all three in one regression model. Ok, we go back to the Linear Regression window and enter Social Support, Financial Assets, and Age. Heres the output    When all of the predictor variables are entered the multiple squared correlation goes up to .292 from the value of .285 we got when only two predictors were used. We didnt gain all that much only an additional .7 percent but the question was how to make the proportion of variability accounted for as large as it could be, and thats it. Accounting for 29.2 percent of the variability is as good as we can do. If you look at the t-tests on the right side of the coefficients window, youll notice that only the regression coefficients for Social Support and Financial Assets are significant. The beta weight for Age is only -.095, which carries a significance level of only .573. This tells us that only Social Support and Financial Assets contribute significantly to the regression equation. Automated strategies for selecting predictor variables Ok, to maximize the proportion of variability accounted for, use every predictor variable. However, using every predictor variable may not always be very practical. For example, in the analysis with three predictors described above, Age contributed almost nothing to the regression equation. It probably doesnt make sense to include a predictor in a regression equation if were unable to reject the null hypothesis that the variable contributes nothing! In another research context it might be the case that a given variable might contribute an extra half a percent to the predictive power of a regression model, but maybe it costs $50 to obtain a score on that variable for each person. The expense of obtaining scores for that variable may not be worth the ability to get predicted scores that are trivially more accurate. So, if weve got data from a bunch of potential predictor variables how do we decide which predictors to use and which ones to leave out? How do we account for the most variability using the fewest number of predictor variables? In other words, in most situations were not going to want to use every possible predictor. Were going to want to use the most efficient regression equation possible. We want a regression equation that is lean and mean! We just said that we want a regression equation that is efficient. To generate a rule for selecting predictor variables we need a definition for what it means for a regression equation to be efficient. SPSS provides a number of automated procedures for selecting predictor variables. These procedures differ from each other in terms of the definitions they use for what it means for an equation to be efficient. These definitions translate into rules or algorithms for selecting some variables for inclusion and leaving out others. Im going to describe three of these algorithms for selecting predictor variables, the Forward, Backward, and Stepwise methods. To explain how they work Im going to use a type of diagram known as a Venn Diagram. The Venn diagram is cool because it shows the degree to which the predictor variable overlap with the criterion and, in addition, the degree to which each predictor variable overlaps with each other. Heres a sample Venn Diagram. Ive labeled the variables Y, X1, X2, and X3. Figure X.3  SHAPE \* MERGEFORMAT  Before we talk about the Forward method, lets look at the Venn diagram above. The area of the largest circle represents the total amount of variability in the criterion variable Y. The other three circles represent the amounts of variability in the predictor variables X1, X2, and X3, respectively. The numbers represent proportions of variability. So the number .65 tells us that 65% of the variability in scores for Y is not overlapped or accounted for by any other three predictors. The other numbers correspond to proportions of variability in regions in Y that are overlapped by one or more predictor variables. In other words, these numbers can be interpreted in squared correlation units. Lets say we want to determine the proportion of overlap between X1 and Y. There are three regions in the diagram that contribute to the overlap of X1 with Y. These are the regions with the numbers .08, .07, and .03. They add up to a total of .18 which tells us that X1 by itself accounts for 18% of the variability in Y. The diagram also shows us the degree to which any two or three predictor variables overlap with each other. For example, There are two regions that comprise the overlap between X1 and X2. These are the regions with the proportions of variability of .07 and .03. Now lets use a Venn diagram to illustrate the logic that SPSS goes through in executing the Forward method. The Forward method for selecting predictor variables The goal of all three algorithms for selecting predictors is to account for the most variability in the criterion (Y) with the fewest predictor variables. That means we want to pick individual predictors that have the most overlap with Y and, at the same time, have the least amount of overlap with each other. We dont want overlap among the predictors because accounting for the same variability twice (redundant predictors) doesnt help the equation at all. The Forward method starts when the researcher selects a set of variables for SPSS to pick from. All the researcher has to do is pick the variables and then click OK. SPSS does the rest. The first thing SPSS will do is to identify the predictor variable that would do the best job all by itself in predicting scores for Y. In other words, SPSS will find the predictor variable that has highest squared correlation with Y. Go back and take a look at our Venn diagram. Which of the three predictors has the highest squared correlation with Y? This is the same thing as asking which of the predictors has the greatest amount of overlap with Y. By adding up the various regions that correspond to each predictors overlap with Y we get: R2(Y*X1) = .18 R2(Y*X2) = .21 R2(Y*X3) = .17 This tells us that if the researcher were only allowed to use one predictor variable they ought to use X2. This is because it accounts for a larger percentage of the variability in Y than either of the other two predictors (21%). So, at this point SPSS will test this proportion of variability to see if its significant. If its not, the whole process ends and SPSS concludes that none of the predictors are worthy anything. After all, if the very best predictor of the bunch fails to account for a significant amount of variability then none of the other predictors will either. If this happens the researcher should probably make the L sign over their forehead and go home a sulk for a while. When they come back, the only thing the researcher can do then is to go find a different set of predictors that are more strongly correlated with the criterion. But hopefully it wont come to this and SPSS will report that the contribution of this first variable is significant. Now, when SPSS tests this squared correlation of .21 the question of whether .21 is good enough to be significant will be determined by the sample size of the study. With a decent sample size, .21 will be good enough. With a small sample size, it wont. Because we want to be able to talk about the logic of the procedure lets say that, hypothetically, just for the sake of argument, a variable has to make a contribution of 4% to the equation to consider it statistically significant. In the example from the Venn diagram X2 by itself accounts for 21 percent of the variability in Y. Were saying that it only has to account for 4 percent to consider that contribution significant. Twenty-one percent is clearly greater than four percent so lets say that the contribution of X2 is significant. In the Forward method, once a variable has entered the regression equation it never comes out. So, X2 is in. But the procedure isnt over. Now SPSS will decide whether it wants to add any more predictors to the equation. X2 already covers a fair amount of the territory accounted for by the other predictors. In order to add to the predictive power of the equation we need to think about which of the two remaining predictors would contribute the most to the equation. We need to calculate the unique contributions of both X1 and X3. The unique contribution of a predictor variable represents the proportion of variability that the predictor adds to the regression equation above and beyond the proportion of variability of variability already accounted for by the predictor or predictors already entered into the equation. So whats the unique contribution of X1? Thats the proportion of variability that X1 accounts for in Y that X2 doesnt already account for. From the Venn diagram this is going to be a proportion of variability of .08 or 8%. OK, so whats the unique contribution of X3? Thats the proportion of variability that X3 accounts for in Y that X2 doesnt already account for. From the Venn diagram it looks the unique contribution of X3 is a proportion of variability of .06 or 6%. In the second step of the Forward method, SPSS determines which of the remaining predictors makes the largest unique contribution to the equation and then it tests this amount of variability to see if its significant. X1 makes the largest unique contribution (8%). This percentage of variability is larger than the hypothetical minimum of 4% we said that it takes for a contribution to be significant. Therefore SPSS will accept X1 into the equation to join X2 because it is adding something to the equation that X2 doesnt already do. At this point, if the unique contribution of X1 had not been significant, the process would have stopped at SPSS would decide that the final regression equation would only contain X2. But, the contribution of X1 is significant so now the equation has two predictors: X1 and X2. So now what does the program do? It goes through the same process all over again. It determines which of the remaining predictors makes the largest unique contribution to the equation and it tests this unique contribution to see if its significant. If its not, the process stops and SPSS only uses the predictors its entered up to that point. If the unique contribution is significant then it adds that predictor variable into the equation and it goes on to identify the remaining predictor that makes the largest unique contribution to the equation. Eventually, the overlaps among the predictors become so great that none of the remaining predictors contribute anything significant above and beyond what predictors already in the equation are able to do. This is the point at which the Forward method stops and the point where no more predictors are entered into the equation. At this point the only remaining predictor is X3. Its unique contribution is 6%. SPSS tests this unique contribution and finds that it is significant so it adds X3 into the model, as well. Therefore, using the Forward method, the final regression equation will include all three predictor variables. The strategy of the Forward method is that it starts with no predictors and it adds them one at a time in the order of the variables that make the largest unique contributions to the equation. The Forward method stops when the largest unique contribution of predictors not already in the equation fails to reach significance. In the Forward method, once a variable enters the equation it can never be removed. In a sense, the Forward method is like tenure. Once a faculty member is tenured, you cant get rid of them, no matter if they contribute practically nothing to anyone. The Backward method The Forward method started with no predictors in the model and then added predictors one at a time in the order in which they could account for additional portions of variability in Y. The Forward method stopped when the largest unique contribution among the remaining predictors was not statistically significant. The Backward method starts from the opposite point. The Backward method starts with all of the predictors in the model. In the first step SPSS determines which predictor variable makes the smallest unique contribution in predicting scores for Y. Using the example of our Venn diagram, when all of the variables have been entered, the unique contribution of X1 is a proportion of .08 (8%), the unique contribution of X2 is a proportion of .03 (3%), and the unique contribution of X3 is a proportion of.06 (3%). So, X2 has the smallest unique contribution because it only adds three percent above and beyond what the other two predictors already account for. SPSS tests this weakest unique contribution to see if its significant. If the weakest unique contribution is significant SPSS keeps this variable in the equation and the process stops. The reasoning is that if the weakest contribution is significant then all of the variables must be pulling their own weight. If a unique contribution is not significant then that predictor is removed. We said that just for the sake of illustration, a predictor has to make a contribution of four percent to consider it significant. The three percent contribution of X2 is not significant, so X2 is removed from the equation. OK, X2 gets kicked out because it isnt contributing anything significant. Now weve only got X1 and X2 left in the equation. The next step in the Backward method is to determine the unique contributions of both remaining variables. It has to re-calculate the unique contributions of X1 and X3 because is X2 is no longer overlapping with regions covered by X1 or X3. The only overlap we need to take into account is the overlap between X1 and X3. With X2 out of there, the unique contribution of X1 is a proportion of .15 (.08 + .07) or 15%. The unique contribution of X3 is .14 (.06 + .08) or 14%. SPSS would determine that X1 makes the smaller unique contribution (15%). It will test this smallest unique contribution and find that it is significant. A 15% contribution is greater than our hypothetical minimum to reach significance, so lets say that SPSS would find that X1 makes a significant unique contribution. At this point, the weakest unique contribution is significant so SPSS would stop removing predictors and the final equation would be one that retains X1 and X3 as predictor variables. In summary, the strategy for the Backward method is that it starts with all of the predictors and it then removes predictors one at a time when they fail to contribute a significant amount of unique variability. The Backward method stops when the weakest unique contribution of the variables in the equation is significant. In this respect, the Backward method is analogous to a company that downsizes to try to get the most work done using the fewest number of employees. The Stepwise method All right. One more. This last strategy for selecting predictor variables is perhaps the most widely used. Its referred to as the Stepwise method. The Stepwise method is similar to the Forward method, except in one important respect. The Stepwise method starts out the same as the Forward method with no predictors in the model. In a first step, SPSS identifies the predictor variables that has the largest squared correlation with Y. It then tests this first predictor to see if it makes a significant contribution in predicting scores for Y. In our example, SPSS would again select X2 and enter it into the equation because its contribution of .21 is significant. In Step 2, just like in the Forward method, SPSS will identify the remaining predictor that would make the largest unique contribution to the equation. Just like the Forward method, the Stepwise method will select X1 because its unique contribution (8%) is larger than the unique contribution of X3 (6%). SPSS tests X1s unique contribution of 8% to see if it is significant. If its not, the process stops and the Stepwise method determines that only X2 stays in the equation. If the unique contribution of the new variable, X1, is significant than the Stepwise method adds X1 to the equation. At this point SPSS will ALSO go back and it test the unique contribution of the variable already entered, X2. The thing that makes the Stepwise method different from the Forward method is the fact that the Stepwise method tests the unique contribution of every variable in the equation at every step. This way, theres a way of getting rid of predictor variables that, at some point, are overlapped by the contributions of enough other predictors that it no longer makes a significant contribution to the equation. At this point in step 2, SPSS tests the unique contributions of both X1 (8%) and X2 (11%) and finds that they are both significant. So, for the moment, theyre both in. In Step 3 the Stepwise method evaluates all of the remaining predictors to see which one makes the largest unique contribution to the equation. At this point there is only one remaining predictor, X3. The unique contribution of X3 is 6%. SPSS tests this unique contribution and finds that its significant. Now SPSS goes back and test the unique contributions of both of the predictors its already entered, X1 and X2. The unique contribution of X1 is still significant at 8%, so SPSS keeps X1. However, when both X1 and X3 are in the model the unique contribution of X2 (3%) is no longer significant, so SPSS removes it from the equation. It may seem strange, but the single best predictor is eventually removed because the other predictors together account for almost all of the variability that it does. In summary, the Stepwise method is similar to the Forward method, in that it starts with no predictors and then adds predictor variables one at a time if they make significant unique contributions to the equation. The stepwise method also provides a way to remove predictor variables from a regression model after they had been added in an earlier step. In this respect, the Stepwise method is analogous to Post-Tenure Review at a university. A professor might be tenured, but if they screw up badly enough you can get rid of them! Hierarchical regression: testing theories about relationships among variables The Forward, Backward, and Stepwise methods are automated methods for generating regression equations that are efficient. Theyre great for developing prediction equations that account for the most variability while using the fewest number of predictors. If the goal of your work is prediction, then weve already talked about a great deal of what you need to know. However, many researchers are interested in answering questions that are more theoretical in nature. The questions often concern an idea about how a set of variables are related to each other. In words, a researcher has an idea about what the Venn diagram for a set of variables looks like and they want to use multiple regression to determine what the Venn diagram actually looks like. For example, a researcher might think that Social Support will still account for a significant amount of variability in scores for Quality of Life after Financial Assets has already been entered into the equation. In other words, the investigator is predicting that the relationship between Social Support and Quality of Life is not something that can be accounted for when the Financial Assets of the family are taken into account. In order to answer that type of question we need to have control over the order in which each variable enters a regression equation. SPSS makes it very easy to do this. When the researcher specifies the order in which variables enter a regression equation this is referred to as hierarchical regression. This process is very different from the automated strategies we talked about above. In the Forward, Backward, and Stepwise methods SPSS determined the order in which variables were entered. In hierarchical regression, the researcher determines the order of entry into the equation. Again, in hierarchical regression the goal is not to develop an effective prediction equation. The goal is to test hypothesized relationships among variables. If youre using SPSS, the method that corresponds to hierarchical regression is the Enter method. In the question outlined above, the researcher wants to test the idea that Social Support still makes a significant unique contribution in predicting scores for Quality of Life, even after the predictor variable Financial Assets has already been entered into the equation. To answer this question the researcher needs to first enter Financial Assets in a regression equation first and see how good a job it does all by itself. One statistic that comes in very handy in conducting a hierarchical regression analysis is the R Square Change option. Here, we would tell SPSS to enter Financial Assets in a first block of predictors. Then the researcher would tell SPSS to enter Social Support in a second block of predictors. In its output, SPSS will report two sets of results. It will tell how good a job Financial Assets does all by itself and it will tell you about how the regression equation did when it used both predictors. It will also show you the degree to which the predictive power of the equation improves when that second predictor is added. OK, heres what the output looks like     Most of this output is ready familiar to you. One this thats different is that weve got two sets of results. The first one, corresponding to the first Model, contains information about the regression equation when only Financial assets is used a predictor. The second Model reports results for the regression equation that uses both Financial Assets and Social Support as predictors. The second thing thats new is contained in the Model Summary table. You can see in the R Square column that the proportion of variability accounted for when the only predictor variable is financial Assets .129. The value for R Square goes up to .285 when Social Support is added to the equation. If you subtracted .129 from .285 you would arrive at a value of .156. This means that Social Support accounts for an addition 15.6% of the variability that scores for Financial Assets could not account for. This last piece of information is also provided in the R Square Change column of the Model Summary. An R Square Change indicates the amount of change in the proportion of variability accounted for when an additional block of predictors (in this case one predictor variable) is added to the equation. The R Square Change for the second model is the same number we just came up with, .156. In this case, the R Square Change represents the unique contribution made by Social Support when Financial Assets has already been entered into the equation. The F-ratio displayed just to the right of the R Square Change is testing this proportion of variability this unique contribution to see if it is statistically significant. In this case the F-ratio is 6.992 and its significance level is .013, indicating that Social Support accounts for a significant proportion of variability in score for Quality of Life, above and beyond the variability accounted for by Financial Assets. Generating a Venn diagram with two predictor variables Ok, we now know that the unique contribution of Social Support, above and beyond that of Financial Assets is 15.5%. In all. When both Social Support and Financial Assets are used as predictors the equation accounts for 28.5%. What more would we need to do to generate a Venn diagram that shows us the relationships of each predictor variable with the criterion and the degree to which the predictor variables overlap with each other. It turns out that theres only one more analysis we need to do. Lets draw a sample Venn diagram that shows us the information we have and the information we need. Figure X.4  SHAPE \* MERGEFORMAT  This diagram shows us that we need to know the unique contributions for the two variables and the degree of overlap of the two predictors in accounting for the criterion. We already know the unique contribution for Social Support. Its a proportion of .156 or a percentage of 15.6%. Ive added this in below Figure X.5  SHAPE \* MERGEFORMAT  Now we need to go get the unique contribution of the other predictor variable, Financial Assets. To do this, we just ask SPSS to do the same thing we did before we just need to reverse the order of entry of the variables. In other, enter Social Support first and Financial Assets second. Heres what the output from SPSS looks like when we do this     When the variable Financial Assets is entered second the R Square Change is .111. This means that Financial Assets accounts for 11.1% of the variability in scores for Quality of Life, above and beyond the variability accounted for by Social Support. In other words, the unique contribution of Financial Assets is 11.1%. Lets put the unique contribution for Financial Assets into our Venn diagram. Figure X.6  SHAPE \* MERGEFORMAT  Now the only thing left is the percentage of variability that corresponds to the overlap. This is easy. We know that when both predictors are used we account for 28.5 percent of the variability. The overlap must be the percentage of variability thats left over when we subtract the unique contributions of both variables from this total percentage of variability accounted for. Doing the math we get Total percentage of Overlap = variability accounted for - Unique for Social Support Unique for Fin. Assets Overlap = 28.5% - 15.6% - 11.1% = 1.8% This tells us that the percentage of overlap of the two predictors in accounting for scores for Quality of Life is 1.8%. Now we can fill in the last thing we need in the Venn diagram. Figure X.7  SHAPE \* MERGEFORMAT  And thats it! One way of thinking about hierarchical regression is that its like flying a plane on manual. You get to say which variables enter the equation and when. The Forward, backward, and Stepwise methods are like flying the plane on autopilot. The computer knows what the rules are and it will follow those rules without the pilot having to think about whats really going on. Both hierarchical regression and the automated strategies address important goals and objectives theyre just not the same goals and objectives.     PAGE  PAGE 12 Thomas W. Pierce 2005 Q  ( 3 : [ \ ]  y r zIL_luw"1IβήΪҞhKhh jhoihYoEh{hD5hDh{h6Yh[rhW656h[rh56hW6hh[rhChG ShS hS hS 6hS hS 6CJaJhH6CJaJ8%>QR C8$a$gdS 8B )7@Ra+9BT-;DVf0>FXv "#1?GYhh;th?LhFhA h.Yhh$hh$hhhYoEh jR:<?@DGIJJK5CDK8FN:HP;IJ;I<JvIt hA 5CJOJQJ\^JaJ/jhghA 5CJOJQJU\^JaJh _h jhoihdhoi>*h?Lhh;thYoEhhs\DJ ""$$2%3%4%6%7%9%:%4&5&&& 7$8$H$gd~j 7$8$H$gd\< 7$8$H$gdA  !!f!!!"""2"T"e"""" ####$$$$$ % %1%2%3%4%5%6%7%8%9%:%쨗meh\<OJQJ/j hghK5CJOJQJU\^JaJ#h~jB*CJOJQJ^JaJph h\<5CJOJQJ\^JaJ/jhghK5CJOJQJU\^JaJh~jh x_h>^hd56h>^hd6h _hd56h[rhA h>^hdh jhA OJQJ&:%@%T%Z%p%%%%3&4&5&B&i&&&&'' '!'V'W''%((((((())()>)B)))+*,*****+\+_++Ƕ|v hH i0Jh3mh2LThi#hi#hi#>*h6i/hH ih%h _h#h x_B*CJOJQJ^JaJph h]5CJOJQJ\^JaJ/j&hghK5CJOJQJU\^JaJh] h\<5hh x_5h~jhm2h x_h\<.&'' '!'((((**++++//|2}2h6i6)7*7,7-7:: 7$8$H$gdi# 7$8$H$gd x_ 7$8$H$gd]++++,,O----.h.m../050Q0;1[1112{2|22333h6i6667(7)7*7+7ʾʕʕq/jhghK5CJOJQJU\^JaJhGhh5hhSqhSq>*hSqhSq5 hm256hSqhSq56hm2hSqhSqhi#5hi#hi#OJQJ hi#5CJOJQJ\^JaJ/jhghK5CJOJQJU\^JaJ$+7,7-7y77::3<7<Q<o<p<q<=!=====x>z>>> ? ? ??4?R?T???5@@@@@@@@@A4AAAAԸЛ{{w{sh6Jh<hhs h0J hv!0Jhv!hhH*hhC=OJQJ hC=5CJOJQJ\^JaJ/j hghK5CJOJQJU\^JaJh/hC=hC=hC=5hm2hGhi#&hGhi#5CJOJQJ\^JaJ-:p<q<===== ? ?[?\?AAAABB=B>BCCCD9D:DDDG 7$8$H$gdC=AAAAAAAAAAAABBBBB;BBmBnBCCCCC,D9D:DDDD+EEEFFGGuGyGRHqHHIIIIJJKK0K2K6KKKKLL罳罤hW.h%ZH*hW.h%Zh%ZH* h<56h%Zh%Z56h%ZhRhR56hRhX56h<hXhs>* hv!0JhRhv!hXhshshsH*<GGJJ4K6KLLMMPPSSSSSSSSSSSSSSSSS 7$8$H$gd'`L LLLLLLLLMKMMN5PNPPPP(QuQQ RSSSSSSSSSSvTwTTTTTTTTUзЫЫxnhh56h11OJQJh+CJOJQJ^JaJ)js1hghCJOJQJU^JaJhKOJQJh+h'`OJQJ)j&hgh>CJOJQJU^JaJhKh>h'`h~MTh~MT6h~MThfhW.hW.>*hGhW.h<h%Z)SwTxTTTTTTFWGW~WWXXZZzZ{Z}ZZZZZ]]]] 7$8$H$gdA 7$8$H$gd11 7$8$H$gd+UVVVWWDWEWGW~WWWWWWXXX;YCYfYkZyZzZ{Z|Z}Z~ZZǽιx`O hA 5CJOJQJ\^JaJ/j'EhghK5CJOJQJU\^JaJ&hA h115CJOJQJ\^JaJ/j@hghK5CJOJQJU\^JaJ h115CJOJQJ\^JaJh x_hA h11h1156 h<56h11hW. h116 h?6b6h?6bh?6b6h?6bh?6b>*h+h<h?6bhZZZZZ!\]]]]g^n^^^^^^______`M``aab*b~bbbbcccccdHdwddddd$eEeqeeǻǷdzǯhzhzhh5hhhyhXhX6hXhX56h<hXhh}7hYhY>*h11hYhA hA 5CJOJQJ\^JaJ/jJhghK5CJOJQJU\^JaJhA OJQJ1]aabbddfffffgggii l lylzl{l|lllnn=o>o`qee]ftf|f}ffffffffffffgggggghhhh#h$h3hhhiijjjjkkkkkk l lPlXlzl|l}l뵭멡ۉh^Wh@Th<hNH*hNh#zh<hCJH*hCJh<hrH*hrj"QhkEUjh;lUmHnHuh;ljh;lUhKhXh<h/H*h/h<hzhzhz52}llnnn`onoooFqGq_q`qbqcqgqhqrqsqwqxqqqqqqqqqr1rFrerrrrrrrsstGtIt_t~t(uhi hi hi 6h11h,L.h"WhQhB5h-h\h-H*E34KL'(=>ݖޖ56ߛkl34=MOklڐef)*01œ&'()=RДiޖ:ab/aƾƶƲƲڲڲh\hH*hh\h4UH*h4Uh4U5h4Uh'@!h'@!6h<hL @h&)>h\h'@!H*h'@!hC> hC> 0Jh\hP}H*hP}C45|}~tŜƜyz̝͝^jlmn@Fɢܢޢ8bd~٤9޹ⱹhvhh9Chg hhMh?(h?(>*h?(h'@!h"WhL @56H*hL @hL @56h"WhL @H*h&)>hL @h\hH*h4Uhh56h<mnݢޢcd¦æ9: 7$8$H$gd+G1Hp789W_:;޼s[UUO hms0J hy0J/jchgh5CJOJQJU\^JaJ/j\hgh5CJOJQJU\^JaJ/j\Vhgh+G5CJOJQJU\^JaJh+GOJQJ h+G5CJOJQJ\^JaJ/jQhgh5CJOJQJU\^JaJh+Gh+G0J>* h+G0JhShS0J5 hS0J h9C0J hg h0JBC;<_` EF)*ɷʷ 7$8$H$gd+G\]_ EqķŷƷǷȷɷʷt -./0123¼®Ҽққ̛̛}pj6kh h 0JUjh UmHnHu h 0Jjh 0JU h0Jjjhkhk0JUjhkUmHnHu hk0Jjhk0JU hK0J h<0JhrBhrB0J>* hrB0J hy0J hms0Jhmshms0J56% 23*+67TU|}a 7$8$H$gdк*78OPQRSTζ֞ֆրztjdjVIjCzt9h6=0JCJaJ hk0Jjh>:hK0JUjh>:UmHnHu h>:0Jjh>:0JU hK0J h6=0J h0J/j ~hgh5CJOJQJU\^JaJ/jvhgh5CJOJQJU\^JaJ/jsphgh5CJOJQJU\^JaJhOJQJ h5CJOJQJ\^JaJ/jkhgh5CJOJQJU\^JaJ&*,|}amnoƺƴƪƴ{w{w{w{wmgmgwm h/w0Jjh/w0JUh/wjh/wU h,s0JjKh{#hK0JUjh{#UmHnHu h{#0Jjh{#0JU hK0J h<0J h=0J h6=0Jh=h9e0Jh6=h6=0JCJaJh=0JCJaJh6=0JCJaJh9e0JCJaJ(abmnh]hgd &`#$gd hK0Jh/h/why0JmHnHujh/w0JU h/w0J5 01h:p/ =!"#$% Dd f!g0  # A2:"N@8$sdQPD`!"N@8$sdQP':)RP*+xڭOlGw6 14iOR8`9(Y!.JzSޒ1K-GB+*U HLN("9^Λ~},N[! ,>Cg+D/ԇg.=ܗb,9yه}@1&Ng봌;gk<3.1`o- BpVsmD=ٹz`&c:[1r '_"!Z8}?YX\ ⏟!-#>љz!\>i& lIOH,ΡsxW*CCl@C,5X@P`c(H^XUAχa :K\O(PKX/Ga , uC$ԵH VY:4VB݀5B,(į%Tdh݀5B,x`fBUP gD, BUVXX jZ$];F0(С j,T% 5 FF+j -zrVm P ZxP,էRR3(ԬCMXC X.D&=|0@jGfZ -5hYcy!V[*> 2 S)$RR[>[{)3"%-i#3&IICKҐaI&AJ #"IB!)oX` HRPLHن.?2#1#9t>)GJj:)fOxw\搧z]AX~'7XZx ճ:4 b+'bK#Weej8tԄ3v"BMjgT9ՎnX!gnwMÊ'B?䔸m޷-g]OI}2b*䈷5sn긍3vUnxu&жtXVRnCz8YGqRmkʶl/ܻ͖mlۓm{TƂ_w~"~_;==`y\2Fϐ&;+twS.r_esDQ"æAPs$XK8x(S F1rfʼ\Wk82dbkq6+T`qn V`'1 ^"$/+D0܋q#{C:SGH;dtЈJĨnd+Kt211:G]qJ%/She{PnrE6R hPNQ!!VXy6u2۞}d~g3%kɐ<*+p$d&m&%gtB5I8uO^vM*b錬>-Ube_?cU9}@aHzajj*9;)j|2ڍ[r\k.҅E՜<;qncDma[ɶ606"Sjskt4?h.nj7d}Z3ҭ{V4f*kؕԕ;н'-<o\~A-1AƗ.W GsWE}?޲@Dd g  0  # A2Hnv2X0* `!Hnv2X0)^Pxb^xڥXoEod5RbD7qbCIA\@Y#5rqUur±āSś?ZBBH@}ABP@z|;^a]e͛7|훝p6X1tv9cp?*cEwlJD`Kt^ opo"Mx_XDp/(srD8ָ+eEWp#%\e }??JXw1s:8i/zGZg]ʈtb%ktG"y6H$cuih4ݱ?BU0Z }v(sr̘ĿMX ( f6Tmzq0 &w0<s|L#Vư \ ZGq6y&纉,1״vV}Yan 7 dR)5AӤP] sfFo\5&4YS厃 mfTox-6|-6:Ms33LCV4_@ _MhYSQ ۴:o8uLδEcG"بo\+)Uw08p$&p)W$ZT_z],Y>.q%dVT9xʵ0o=9~ Yex[|]ڔຬmTj#^Z8:xpuylA ](ߎBZ(}i/6b1TXk;-^nzcլZ^-qʝG;"pċ<_ɛl,񳢣 K..0#ttHF$D-?!z[·N_Oa rk{&D' Ǽ`^OvW:|;yw{ QOKd@rkZ+/I;ɏ˥dTFVu][4"kݢ3Ly#ϝYU>uqZ]n29>}ȫWRa[W2=k08j`!}&C#>8RM`JxڽXMlE;@H-kGTMROQ.z\;zh8p@&"TrX₸VpPKHqTEz|;$e)vw͛y|sƘY/~\>ʚl.uoICwGx^_9A gw/-xIo>q>=R52zD<",86JQX4^g7+W`{k$;}ڥ0oƽ WN&6XB>Oy`ybxiu(5: ~v(SzaR >K_.t߽aI,ݡM 7NñsQ1 =?CHr|"1ˈ?os)9@/_F9B!0,՚G/0jchCGaBbCj&}/Wg?u1*FqK{9IEEݯRtx$^F3KP~R%(b1ˈ߸P~T~ _.-U~^qDI[qWY(kbpd4bWY( (_F#`{ qGO@=whU漵j)"NE;]:.Q73K+̜ͦWGQUȦA獠Pyp($P,҄Kug~ 3o40]~J2].`hc`n(*HK@;\~nY5}箓WlZ j. 5;yZneͫvgvbWbC%f"PThg-3ܵ&Djg.wZ'\VIBpv˞:6KĶKNɶmd-6\[q=W/QZ4Wz%Oi7[O)2g&SWs\b׬L6^ĭ\Ϋg fт"Mp@S5mTFXz^K!5Y343dy1֥T&2RԴi5ь 8)CըWP-xNЎZ7\-$m)-2;2ql*'(ma{9up!ٜy1ɰ^aKnWސ9OZ <Dd kj0   # A23QO%n~ 74D`!3QO%n~ 74Dh u7ZxڝOAgK Z=!fm4 R XCBh!B/@6b⅘x$ҳƃ'1yn[kXL͛̾Ck scsN*%0 Bd -^!j+Pmb{ ;sC_ӆ6# ̂I kE?9 dPf~Iix6i~_g2iɤ m? >R*$b;u.A>C.z:>qb].Deb!8164j9xy2Ơ#kĐ?;]C[Z85}a& 2T,-Fp2 \`|) P=@e BcQo>DHզ)m~Ա ]R+hJDp20I+K@9C08] ;m/Yq 9s1ӚUCKgY;UѠ8NQALZꌭe\CF5m0K}fX3%j4*)A:KM6ړ4IYIrl^ *b: {-ӽzQC|maL!I/كWtfx,1)jt㾒JkrJ9,=H,R$ka)n4],JOHm&hHՠ@c96q~_9͸A g/Mx_S~%"r}4}ra{إi#ɨqsT? x&O0޾g'+9//KvK!ވ{;7 @mmh'] ۞=1XMqe[mU!Xp!`x!pYK#NhhhtXx8C[rB|"Qˈp9G/ p(Ű#`M5^6M È6Q:0 ] -)4885_Ư?u1*F\s`d$CZw \MW%I-yb5gTΜ3ΛWR%r3YYp4IS/QPLM}{.!}G]P9A3BM wF[(ۢSWe[tm~ejule:`UQiSs퐵;p![ٜu1ɰe_KnWސ^NcDd ! RR0  # A2B,c6l \܋+f !`!,c6l \܋+f5RxڥOGgNcR$ʥC?)V4D9R9162`|'DQXꥷK/ϭrȥi 4$ߝ]/6]Gzg߾y7ߙ?]`Ϗ) bb%7:<"Mi+8q3Ľ38?7Q""LJ>vE ,;FPJ: Y1~vd)ɣ6M~h=xrg;? .~k1'9it~pDmhJ<3 d+_X@G3nv#tdu9A +ZbnW,% [q=0gu:t8%n2aSDc `j ۭŽ?c[ Ϛu0`3L?0%IjE'!XRu5nHt\`Tz⚣JQ}<g )QAkkТ-hI-K-Vk.VLvM;]F;U,3{s.fQ/CcAQ*o'4{yX*.Əaɲ ARK>2\Y1G%$̓P!v!lŬWI!#[%CK:0?Us5~~ `RYP(Skr^n`^ޤrV])%5kJ(فf*j6S2֗O>6_>*YFTג3鮽4׌rF5_ȗsvʯiD\̻dӣl W{ePm q!ﭔ%]soM0Na Dd $@H  c $h7v";1A"I 7Zk`X"r% &@= 7Zk`X"r$DpZH x\ pTW>owIB~!Oa+h$EJ;ԎH $m*?RR1S8èuѱ*ZQNSIfBwow&t3g߻w==G5XgQշY4O4OxGlR䘥T%w=ʵ:y&|``f(ri3\9MS)fhI { 9RP= hz>W~@ :΂~jUOz;hn- :j-@q1Uj6P33L ^jBylj.mWCZjZ)-3m⡟xx79sKunc,X!T7.K[N߱Aoj2,tU/w>,>|RsLDervLRd|T=Ir)J- tʉlA6K/gYaޗ~͟9p@)3`g#ce|O{~RusPчjAޖ5?Mss x1xkϩ_ǓV|zauzo^Ŕn ݟ;x@nNש? 6QqKPVTzp-)5spSEu нhܝ= xs+͌+ob2^CW F)[vo'e Eԡl/17g e.pP\6Fwx:W{6藠wԥ; 60ׂ!%O/K_Mg@yb P.3x =)`,Rys&N| n_͠OcDiDasq$0ʿ1?=!'Ϣk ;ns_(wcY<}CON#JR1g}Cwݕ (5^b 6i> (gJ!9^"ieikɶcۛ~2lopyE8knH>K.ЅzC_gxx_ kj_v9tzgnHߢrɲ*|Yg}_ gpN=em-q5z #~Ӥ-\8BKmk :.;{(xyHٟ_1|^ʡ+0]xu_:d:uB ]ZӨcK[?iP)J : QL=8/.Xچ>+66=ݽ133AΛͽ3%ƒXAkA_l ^ls<xtPrȚd+Ab`Lxo1P ssa1Ɋ e#+Ed-W7(֥*"͢XdEuآʄ(d1.#>$YS85yDN}Q7>$YOtϲ>nv5@?ەhvo"pQ[zM1^P'6xg\},CYYSܫɸ 2K|H2~_'feM??_b_sF>^yPoV+톸 .nd=yQewec/&D^!61m3E+xt9f'}JdnL]7,^2*!T 2)(mߦƷMZ|K;9΅=){k1͋{aĶFw5pi1F]NܫSk!qA_=-C7yti;e5goͲD.+?7~l'-K4食G;2~[$AM;G GTw *CUFZQ|x"'iݳn!S߈f\V7ޜʹKuRz ,u5k] 珎g6}f#x&v֣{#u>m1fFg(ᅭ(XMpc~_m˭l2Ik.?WנEȕgGL$2?S#lDd $@TiH  c $@17v";1A"^r-*.^1@=^r-*.^wpZHrx] pU>7 d!f#!{!D&[C%DGP$e H, R0(Ep8ZJ8CUV %Ὓss9}BES$ķF+>PBdN|h"bm"D=,.a$ ,V>GTwuqt<>d>niiIR| E|HO;ϑRl.롵W^QwbwCl:!kI~8>HmWn(ÿkD1L^**Esco' Q㞥@|o9۽2\?OUƼg )|5?Џ?KcGpw}\,&1X',|~KPK=|M&H9C`c?8Ht@'؅ l#b ?GD:Pd i95lei} 3ˮ!n>C=$#J29c]Kٟs$}>#mq =IwVb pPCmuib2gGj&um'\%_?K3gķpKbOJ97 _V~U?Gp/5ȶ;˿K#ǟ٩ŸoR>`އh5X*3w+y``Fc;]V[Mx%uj׮qi}c^>U:t˩Yc]*'gO 5>'c]߇{ iyoEw4K 6MZ;r;I:2Zb7?\W+xO7赮vS5o|vn-w5t'(>{V?k⿉g}fϫv:$/%aF*} z/BF-_ZK`%b.)6 ݋khHj >`܂Rn{@;۶9u.uNgS5RԐ 0j#|Y}!&t <3n0ׯ7 tsܝQ?rp,Ec*XiӠcZ{9Q4sGc ӰN5Vr\&`<~< l/\? x~ngؖ92w71VN0/s;?Ws,ƇC7;倝ն~z(n<ƽ1yvEx4N~ŀ s@?ƺTsli5TQ F7pxN#k4k2ocu=Gpm 06 7_j. 6o:X*E?\!^0ɰq1R?>9F9g+ꡪFIRNlsRںKO=:o!inaza㪣Y,D֩@ZTB)ėT$=?ZϞqQ/;6ޯyB0d({K"yEiWz';abϴcZ^VϴbZ Zb}5gZ-L+hOkk/}dHwߠy&\3"!IJ$^3oϯy}Vq4[_\#X!ǬiѮ{I뛴ůjs<<6j-yL1L[ZmMfGKa!/3켛lK4V6tg?KE&oMciBìt2mGI<X\{/p<ւq?x+sNvZoWza_ȠS/}NӘ:&ٙ;ˡ+ٹ2kΌK/}+NOk?Hs }b NU7ߺ qMV*!(#7¶ׂ zT!c"u79HT6p1}% >M$Ssf1G-9~Xiχ]Iv&|boa -OV;g> y<x$EGDJNX&WOVŬtL7{.9B$83 _HD{C4@31r8gh^ ͚z/0vŊg b( 1t)(A{Wv L .Q~w|>9,Mcۛ͟/>Fyءy2\ yd+գ}:VϘЧM{~d1cG~86js3a%8x IzOJ$;"⽥."00GC _{hs4-}K3c K,LaŜ|0dfv_opBl˽S3Gvp=y9 v\/;3x ZІ94۾&k 1? 3iE'q)my$<~I8 x| x$|wَ~ka~7~y!s7^Pm$:/,G[ ͍<;z慝{M1΅&Dp179jG_R̭iG[̏7kC:蹊qVC#;}뺞qtC:e"zט=^IW2o|^0}:cgHDd &yw0  # A2 Ct-!Qd#A`! Ct-!Qd z6TfxڝAOAgҥT6P5*V7`CTi[Hip0xJx1B((priqeRYXKh5:uJ5[j2eu$ :攠9A'X$D+HLLLF' `gC/#(cԀjTФL1NXx+@u2T`Tq-9k?V61Ǖ3iͪՀӼRiqE(X#B<8R&cYfz~f_LZ2"MJ8*np@▄ āS!foqXݙog֜0Vǜ_W7<[cqJm+vBPznG8Obf\ gw&>L>6KEz*t,}ƟDk1Tap`(5 [#/aW~M1z$k4֍\?6prKƷQm6: ?I0 [ё1m Bu`hw1t 0CX#+vO \WÙ.b$^J>`?&q&tBȘ5<6f"/KP Et/!~/B R%(F+1] h%muǹ |ve-{*@Y ӰhC(kSVaK?plo"N*՜f8EĩhqJǥ>&NFAv|}=3ggLZwU"wՃC5h8J|";y=(T\ 4%R3u|3v2u^kLdl] 7yY \bE2<zq#uOU2篚c.?wݼh>ɦtv^PWȫe֍yՉ*w*vlNMEvDy^?_S{l9m&zEU;#wx =ZIzHXکo]ķnxFlkiǦJ-mݳ1Ej^mT:J)Esf)I\f.{|L:Ks/%ʡTꭽKáRH=ҴήLYGZ<;;}}7Y;c46n.C=F#g!B뭺>Vpm8Θw'SHgGhíkɅaq}_)*IkяxNfl0KShޯm|G7{8X\3ĉ;9i|<^1Er5O"Y}q1y'g+f1~SW?6%q1kAfܜ+^iFX5zʴs ׂOߠ'l:Bqz3LG #p G`M *3Tj olC̳ŝ3y[: L.Z 0D:q5\[jڳun\J@`D`yFpJ*A1$!˶A bs/^,GƘ*CTZh<Xt.IPۥb1!41Ë-8`;G_Y ;`;G 8R%8X,؎,‘)Ñ YY#S9G9Taɮi8r|~z87[p48o0JPA;WB YTa"!FoPc~4Bz~L%ۏiO"4X6h5E/96<~oR nv Ylf%?PIN8%:8:%G 4nl.6 KdYKd)%ESL<#q++/b=;1KҤA3OỺ,S|7jۤلR^a8Lsi7㒡 y dU5xͪӞLw4TWkCN嚬 UԆM'|\6-8`ӟ\;4LʷE"K$^@11TT@5zi2=gxjz>Wz9!d@ĐeZ6P>c;r_e Y^@:T)WݓOϪ{x8s?O;'#zEmz[=bA_avˈo_ҶW5)o}u tu6u}@>EH.gV<үx?O%E4U4( M) ǙwI>8q ɳ6)̳ DF))Yh+4/"̾yu2n\iƌ}GcmTϺ=N-1DdVQD  3 @@"?Dd  pn0  # A 2&jl} [ &MQ`!jl} [ &M: ,К+ .xڥWOA-Ϡ1xm 6&*b$)l. &Vh/x!hZ1H"XwH Gi7o|k[Jj˴jaJVRiȂv| y mSQ5 F9mI|<^5ɏ9@#Ϝ=b {"Ԃm:dp=f~/l~qB=9so(uH|x_DU#7JW'5ŗq5% nY%DEaN 8>r C8qYs-Fd^QK"^Ru vST 5ZntRb,FI򺮡4}xZvu]뚢w#RǨ]Uꪻð!:l ldrx]Qm3*x& c$in"ϟ({nV}ھqfi?)h5ҋEKcp Se,ŷ5 ˺S%c|OeiOQ?-RӺP-8):!̔lYӧ8;2W{;30}v0* 5,w "SL]İ<bxMbد0t5?W5tqc93W՘:!6BG_Rn.ƊM+fU;6Gw1rZ(a"+|+}`f&2Yb++۲de %mY. B^!ۦӒ4+2V`!+##p2wcs>hn0xڥMhGgfVl'VEbn KUBR҆\6XZGrCk\A JiMI !ʰyy3CI'!Ƴ0m A{Qbߓ@}]lm߁5k'9߳mIZڋQ+x e QTζ.5n1d>c-@hatg̑閭\g$>E)JW<ߠݠ]k Kfj%zh< >s|EPbfc1g,-b$md&Ʀ\:nSCoHQ h-ZuKu(O钜hTl#q]5 Q}8K xlb\{F* g2l&Xo Rmd gNNN2p3ty טM0ĢՂo ŖVu۪UQ"߃nP#$+4Jp"$Q%awO?bygSEoH$Y-!F|];Nn qsӤuiqhlyZY^Šqv q7,V%ȽF4ޝųS{'Ί/fv1kCj'T]^i]M [WW'x7t}<NwxgSd267,!SKehGk))2h?#Dd J0  # A2%rF~[^:o]`!g%rF~[^: Mz$5xڽYMlEٴii5$7R%8T9&NeemQKbqDnDNU.9 B ۟LbW5ggg}oޛ(1c] 3>n\.yōto܅қZ ߖ+ ;p=11n qJD$(v9FSqB|Ժ-zqmd=O%t+cӛ?mka{dwL)mTPlC1.;!;dZ٢sfFT'JwWʋUC!t%.XGXlBh $Jmr l &AS)0*ky;| aۊS6ѻM1`k4^  Ia{#5zl_Wa[6L`̗ݷ/C+k]ͰRnB%(/C+ _}/ASe("Z`^&z*eOeiuYV( ʺ# e-Q(#]ŰڄeBY@ C-`{v-3W[9H=֌\h,aЄvE>.@;:#ayTZS8"#6ѻ|\@>4^h"/|\@ԑlr^-jnC^/ NQ0FFej N79Pa߹D^MCSPhHvU-14݂8śȫ18uWǡxʊ ešt^-ȫPOz-P=Zc5(i;%ةJ1JGM}ThE5ӟ$'sT6#]Þ kG +K,$WkX?:Aa(Vb>l]:ĹGG ;H 8MxDIǔCE L ā"#h sQu01럹*av&)a`PL U`X@VcГsK~ ؓͦ?}{QC6Bڗ=b羝ig/$0etYJM:v fg2mffe/þl5bL2p5d"㳨]8g[s?V($*InҶbCj[@Z~֩{Tl*ߙ-QVݎ> dY>3ꬒQWuEQ!Sa0A X;A{.Dvؗ p+ət>;;^6'2}T&N%tX.8XD}{}"&OƁyMfɅ7yi7 kҼ@4oVx;en'?RF23}W5=y1V-w٤cmyG^pN 8mramgUnu8hEv%T9TQϭ]ۮ&NTP]WvCQ{Ji5.*Mi6eٗ&dt_5(D.LDd !? 0  # A 24UZ,읳8҆z6^C'eSj?(FC 8'ku68^L'(rX^popn\ ȝLz2N6dd!NKˮJOmt  uJ~Lac_ʷ>'|m0pbп@]lA r_5GFUaVAC5tyAm@ HdHUA bTMiWq6ֽ&zy; ͑!ʪ)(s&mY Vl + \`VybYYZݚV,&4XÖ(K`1.YLE'!E6c$iɷܠ:Yx cK<1Y/қm#ut G@*Nx //'Wq}J&8t|x%+y^^fjL󹂐2wkiP,ΏqCc~Y{Ş@/ ~*ANo*4TOrx/WeLYPK^vG1zyL(?KwWn9K`@ 1l"5~$TB ]v`"*ģdx?A1ڃ~{of "-p C=[GHK`X*#8jmO w^B p~'"6LHG)#E_O=X պ4~)[(bRd9HPU/|kZٜTӑȂnGjJ8>N6uSO]ozRzVpCsh< yֵ7XXhG7c~#]!]nQ?e=ig CyH;-1#0vE-KJʮ5u }eWf&F}㈱.*V6K5Ee}bxq(x.«#nO4S[;.({״< cSj>k=F+͏j0gzFO' ' ԼYU N{p$V6Q6A|g:ߪ߆I$ tR!{U{UW7ĄΔR]*v9lJ:?&j[c'-߼]^ƽ CpfW!7 1(}Dd ," 0 % # A2cC* 5H~Up`!cC* 5H~Uhn8xڥMlGgf| Q6 '!ZA}H,N 7ئ[U/CJU KUUc1EWTUUQ î2HϾ}37of;tb|{09I_A;H~3a. y bS|(\XvM;~W_\Pǫ(ۤLxRw,Gmn15{6f^c3u~/o=eu~IMQCnP63^f򲕿=۬})§З޾4NnҬ/Ee_U$DAqx< |6~mF"]2F1(bz1ts 5zuSu O뒌 jq0A*@p I]5P}^SRA"H%.S_Da|O=AT*D6Ap3dbL!~13.0cB/C| Ht2`b%bTVTiAJJ͓ iAJJf3}9A!m}%e}43x'e j`e(@M_Sć"غcpIzT hCPs=@^/#|{]]V~Qbִw:}ZRl-On >AwzA+%P 1|>oMlGzv).- UԦ}J{ΌnTs |֟.YkڄyHM0ee>gȀBwQ*\S̐PP/\/TQxJʺ%eŠ%hSh>B=yGV䂬y'I b|Y8$)Vγ҂yd$ pHF]I`hHCQd=ΣV&Q}C;[WZ}/԰N@엂~')pIP~)(xUK\Is7VŰfJS ȌPX:!"(CYRYQ(+\WGh(mzFhsVY+ƭUO~~aS%cq|OWl"V kGc8#%UwM?:Aaeh2Q"}ŒUtYCqD&#UGD2")tQI(PDE`-jbf3WwX>LU1,CYy PEgT0=9(?Bj =j!c#ݵ3*%Oy23ܵ+}| K&+ʕC2TfVPiK4= U+gUJ㱴ǢBvm`w7X<2}\;sHRWgU`/Uh6{*WQzeQ1f 3f [A v_ڪ޷^0Fm5X;kT\]M㹌;]O'r9[f\,{3JĒW3sZ3*V.Ps[@RLմ:ׯ̒5 KM'RdN쓥~O$MH6eKtv.+]ʼY;|. t<\%bsHn {O3N9uugnclw]zg#gK%bO.i9ByC^a"vkw9iWvVS!G==xZ#mߕ6n[B)tbC?ž0&cG{RDd !? 0 * # A2+X\GZQ*P~`!X\GZQ*x5R`xڵoEg&nl(Mr Z@N"ƱEfYWm9>#!?@Oi! 7^( xyvv=:"Ƴc :z:tTit xׯݵqsTOjВ7I: >|W>)@nj/L@KHK2-RK ZE)gl~>*̎1`v1A}rj ]-f10oxۛ9}K hR;"c˼j'Kcmj=,_G-0۟ke rKX+Wh2 N˥oc,Y']HƊty6Yf=ۄMq:^[t5薷ob}Ǩe*i]_& {%&|1[-xkursɳUުdtۏ-}:COCʹg!i1 Qk~ќ*/- c߼Mb]s4]G*`vƶ 5 "5(9023Y08]Φ?g ~x#VnKs+ٹ̽vhuAMPe9gajpʞgg1{eޝev9̚c)*,7.ܼk(]:Иw{ 7oOϑd:oi+bp|%S`|Ez}.=LN}^.{ur?DdppD C 3 @@"?DdD  3 @@"?@@@ NormalCJ_HaJmH sH tH Z@Z RY) Heading 1$<@&5CJ KH OJQJ\^JaJ \@\ RY) Heading 2$<@& 56CJOJQJ\]^JaJV@V RY) Heading 3$<@&5CJOJQJ\^JaJJ@J RY) Heading 4$<@&5CJ\aJN@N RY) Heading 5 <@&56CJ\]aJH@H RY) Heading 6 <@&5CJ\aJDA@D Default Paragraph FontRiR  Table Normal4 l4a (k(No List4@4 Header  !.)@.  Page Number4 @4 $Footer  !>6@"> RY) List Bullet 2  & F>7@2> RY) List Bullet 3  & F8"8 RY)Caption5CJ\aJV>@RV RY)Title$<@&a$5CJ KHOJQJ\^JaJ 2B@b2 RY) Body TextxFJ@rF RY)Subtitle$<@&a$ OJQJ^J !z!z!z z z z z z z z z z z z z z z z z z z z z 6#*/ 7AJMQ]|cn{jÝ cK/N   6 x''@ %>QR C 8 : < ?@DGIJJKJ234679:45 ! ""####''|*}*h.i.)/*/,/-/22p4q455555 7 7[7\79999::=:>:;;;<9<:<<<??qBrBBBCCDDGGJJJJJJJJJJJJJJJJJwKxKKKKKKFNGN~NNOOQQzQ{Q}QQQQQTTTTXXYY[[]]]]]^^^`` c cyczc{c|cccee=f>f`hahphqhhhhh`laloorrttww9{:{g|h|~~~~~~34KL'(=>ݍލ56ߒklmnݙޙcdÝ9:BC;<_` EF)*ɮʮ 23*+67TU|}abmnĸ!!!!!!! !! !! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!}*+!!!!!!!!! !j!:!!!!!!!!/4!j!!!!!!!!!! !!!! !!!!!/! !!!!!!!!?!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!! !! !!!!!!!!!!!!!!!!4%!!!!!!,!! !!!!!!!!!!v:!Z%!YN!!! !!!! !!!!!!!!!!!!9!!!!!!!!!!!!!!!!!!!!!!!! !! !! !!!! !! !!!!!!!!!!!! !!!!!!!!!!!!!!!!!! !!!! !!!!!!!!!!!!!!!!!!!! !!!!!!!!!!v:!v:!:!!5!!!!!!!!!!!!!!!!!!z7,!!!!!!!!!!!!!!!3,!!!!@{!!*!Bm!`!!!!!|&!!!!!v:!!!!!!!!!3,!!!%>QR C 8 : < ?@DGIJJKJ234679:45 ! ""####''|*}*h.i.)/*/,/-/22p4q455555 7 7[7\79999::=:>:;;;<9<:<<<??qBrBBBCCDDGGJJJJJJJJJJJJJJJJJwKxKKKKKKFNGN~NNOOQQzQ{Q}QQQQQTTTTXXYY[[]]]]]^^^`` c cyczc{c|cccee=f>f`hahphqhhhhh`laloorrttww9{:{g|h|~~~~~~34KL'(=>ݍލ56ߒklmnݙޙcdÝ9:BC;<_` EF)*ɮʮ 23*+67TU|}abmnĸŸƸݸ޸00000000000000@0@0@0@0@0@000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000@00@000234679:5 ! ""###*/,/p4q4555559::;;9<DGGJJJJJxKKKOQQzQ{Q}QQQQQhhh:3j00f9ch00eh00c00000000j00JL:ch00Ij00G@70000000000j00<d;ch00;h009r@w0z000000h001(cj0-00h0-0.j0003050050050050050050050j08019?cj0800h080.050050050050050050050j0B01Ch0B00h0B0.j0E0F얉h0E0h0E0000000000j0Q0R<h0Q0j0Q0000000 %%%%=@ :%++7ALUZe}lxiadfhiklnprsuvxy{|~&:GS]`q3abegjmoqtwz}c]^^ĮǮ-07ORn_____ !@!!8z{@cKz(  Ae VH!h+ i3  s"*?f j s *X99jJ?VH!h+s  k  t@#&A*:B/C<DEF11?AA)&;P*(;T*;S*;N*:G*:=*:-*:*:*:):):):):):):):)[;)Y;)Z;)X;)Z;)Z;)\;)f;)q;)|;); *;*;*;$*;/*<6*!<8*6,b;9,];5,X;0,T;*,Q;#,M;,M;,M;,M;+O;+U;+Z;+a;+k;+w;+;+;+;+;+;+;+;+;+;",;9,;H,;Y,;b,;f,;i,{;i,q;i,p;h,p;h,o;d,v;a,~;_,;\,;[,;[,;[,;Y,;_,;b,;d,;g,;-Y:$-z:$-:%-:&-:(-:(- ;(-;;%-U;%-p;%-;)-;)-;)-;(-;(-;-D;-U;-d;-t;-;-;-;-;-;-;.:-:-:-:-:-:-:.*:.@:.V:.m:.:.:.:. ;./;.V;.};.;.;.;.;.<.<.<.<_.;k.;c.;q.;.;.;.;.;.;.;'/S;$/[;'/];+/d;2/o;8/{;A/;J/;S/;`/;c/;g/;j/;/E;/k;/;/;/;/ >5]x\*b1.89DsԃƅQ#3w7-xF\5ۃቛDҐ( Jmϯ  ` l v>BhBw) 5zFq9-4B?ߠ_bjU5B'5<+Xֵ1y4u)&  11$LL'>1  *׀l ==?nlra<zY*q Ɣ޵o[  @@l 0 FFoSyi[dUՅa!36a  ol *7Ǒe&c>gV_|L  匁l EHVHOÎE5 א(HYB#|4Þ#iθt  02l 6 (X#=zG<>Pd//Ŝ@  l 6 (7e א>{ei6Ȅd'^4@  R l 6KHKU'Hd>ufUx[a"e8  +lY l  H@0AX:B6Cf<DE0F11?71J;31F;51D;.1B;)1@;!1B;1D;1H;1T;0^;0q;0;0;0;0;0;1<1<&1;51;:1;;1;=1;61{;,1_;'1O;1C;16; 1*;15;10;2:2:2:2:2:~2:v2v:e2c:L2Y:32X: 2X:2^:1i:1u:1:1:1:1:1;1&;1H;1i;1;1;1;1;1;1;1;1;1;1;1;1;1;1;1;2;2;!2;*2;o3:p3:r3:x3:3:3;39;3_;3;3;3;3;}3;v354<<5,zYqF Bҥ1J.<޵ZT7wdDI+V  ~l V%S S 6f3UFc>cm22!K0JH)N4ӂ<)$L  6ɂl : A<$Q 8<g%3ֆ@AZ@  Kl k4U>FW\Zczsk,  l *7_d`uy  l f-\r\U|ov7b 4ILwh1ڰ| %`%x-H؍eLx\W5(S.'>  {l 6 hs\4x jjl _'}r>bӴ s{ 8RZf݋TvgBG Sm*Z#Ys~ Ѓ9LEEDTsv  -^llJ  m  @A ;B-CIDE Fn11?i#0;8#&;%# ;";v";!J;!t; ;j<%??cY@ AAoBCDKEFiG9NHH*$IkI!I"J#I%I%I&lH6'H'G'G(EY)IDk)Br)BV)kA)@(\@(?j(d?'Y>'==&<&<%i<%E<%;$;V#x;W#y;Y#y;Z#z;+<+<+<+<+<}+<{+*C>+a>&+v>=+>V+>o+>,2=,4=,2=,;=,J=!,X=(,f=5,=D,=T,=d,=r,=,=,=,=,=,S=,y=,=,=,=,>,G>,d>,>,>,>,>,>-<-=-?=-b=-=-=- >->>-r>->->->->- ?- ?- ?48@                   @         @     @    @     "    ? $Ǽ?4|F̛] h FXTO"dFHE#9F#9W 8 e 2 .ZE3 [E8/@?J??U ME_F **7WkZvbqxYN"kuqi^3mo3QXTLTfjڸ&&!xMLc-|Sđ5jLb3F*+sLD³W:fpL'YŲfMb7R7θe9Quh"Qy 8Ƴn*&Q &~6~sPD.<]PbuxL]jtb)¦Bpg(EgJ79b.g1&j`\Rbc3s;]oeu:65gMgULMB%r ^6'fUFu;TԥY\bzGb.i[aPN/t/Yֲ*, ae%, ,,,E RE @YbJ,,X6T,K,ܠYIDRʊJK*XYeBT\d6%.p  pkPul _)@h@o\q̌pƼ vc? ÄپFarŏE_D٫ԥ!)]8rރy'3110& _??  PxIl <AAQS};/|kRm¤ӝg~  ƅl 7DPDag^| F>͘hcvЄ`5;o  %l CNv< zXȰApMu`Ë[ 8c9)c4  `p6ll  D% n  @AMB!CRDEF11?MMMMMMMMMMMMMMMMMMNwNkN^%NR0ND:N8FN-RN#]NjNvNNNNNNNNNNNOOOO O O!O%O O$OOOOOOOOO{OrOfO]OQO=O.O OOOOPP &P4P!BP.LPE]P\lPoPPPPPQ)Q>QSQeQpQQQyQiQ[QD|Q/oQaQQQBQ2Q&Q(Q$Q$QQ QQQQQQQQQQ!Qv&Ql*Qb1QS;Q=JQ8]Q5hQ8sQ@yQPQ[QnQQQQQyQmQ\QIQ7Q!Q QPPPPPPP>Q9Q7Q5Q0Q/Q,Q*Q*Q*Q+Q-Q/Qx2Qn8Qg=Q^CQTLQOVQGdQBiQ>yQ:Q=QHQUQbQuQQQQ=]Q>hQ<lQ=yQ>Q9Q:Q:Q0Q1Q2Q4Q5QzPPPPPPP@mQCeQGeQD\QBWQ8SQ4QQ,NQ!LQMQNQNQRQVQZQ_QfQlQuQQQQQQQQQQQQQ*Q2Q<|QBrQEpQFeQElQDuQE|QGQGQKQOQUQ\Q*PNPqPPPPQ @QaQQQQQPPPPPPPPPPPP|PlP]PYQUQb/Ql@Q{ZQpQQQQQQQQtQ\QBQ,QQQQQQ_QXQaQYQXQWQWQWQWQWQVQTQTQTQRQWQ\QbQnQzRQQQQQQQQQsQpQpQpQpQ R0RIRcR}RRRR R R R R R {R\R}R;R7R2R0R-R*R,R0R2R7R>uRBSRIRNQ\QbQfQwrQiQcQhQpQQQQQQQQRR RxRmRbQXQQQQQQQQQQxQkQcRc%RcuoR]r+Us5q~722M"'HU'8qsCTcکh@`6T  -l <#G$Ұ^y_w,R4|Ć9y5@  P~l i.fX]1\'Y|=8>00071Qu58bPy.7js)! }}{  l ^)$/L35qPfriÚ<+Ƹ)z3k1)>}  ӗl q3&1qzN/Y5uUnIsus~wiY4qa;ǟ{X1C#tL7Lx3(&fa).kLDDUR&6@  0 l :,-Z3LT/ß8ʺqpbieҎyV8kˇ-X)l Շ&j`zַſVHTD'32ɅBD_?L  ˍl c*!2!-utpkT9sU33<'F15n7w< ["3e"bPD0~  l S""χz㋋&1U8yhO69^5s '+Kבshe3)?  @el :JU~uŷ&dyd*mÄ:ֳ×  l 9HdvZXK h9&(mSC(SD(SD(SE(SG(S?(S!)dS )^S )WS)QS)LS)AS)>S ):S)3S(1S(.S(2S(3S(5S(,S8,S1,S+,S",S,S,S,S,S+S+"S+-S+8S+FS+SS+cS+mS+wS ,qS,mS&,gS0,ZS=,PSA,MSD,JSH,GSG,MSG,TSF,ZSD,cSD,nSI,uSN,xSP,ySR,tS,7R,[R,R,R,R,R, S,"S,;S,TS,aS,mS,zS@           @             @     @  @   @           @                @          @    @           @    @   @             @    "'    ? $?4|F̛] > h FXTO"dFHE#9F#9W 8 e 2 .ZE3 [E8/@?=??U MEV_F ]2;p;q L&"U5.1H1E]\L"`UJSSzn<2Kfn20D@LLK>H  `|l t7h. tWaaVFtc~|qF`ǣ]ŇG55 ˎ3\Mx^'ƫSk8s;ϲvҨYRK*,~y  @l <blXtJZ[-88|7qRw@  PKl . sO"oF9Rx  Lxl 3 Xڰ8hX+mۥC6c#  Wl \(6h6k&&.Tqqo0εZiS5pW:%kWfW+&B'H  Гl }A: :fq)LEF1QUK-32M]@848KY~y 1]yWUZcZq〃8wmq10&$S3q3uJ̀  pl X(;\;W狫'QEo1:˜\887|u8֜S9^<~^PW3$!DH  `Pl 9==}UH?779\f:W>qyuzV\9jk^(7~7]¯WNg6yx9q)  0R0R0R0R+0R90RI0S[0Sj0So0%Sp0'So0,S]01SL00S:00S00S0+S/(S/&S/(S/#S0S0S1R1R1R 1R,1R81RC1RK1RS1R\1R_1Rb1Rd1R_1R\1RP1RH1R71R)1R1R1R0R0R0R0S0(S07S 1GS1GS,1FSA1;SQ14S2Q2Q 2Q"2R$2FR$2rR#2R"2R2R2R2S2S2S1R1R1R1R1R1R2R&2R<2RO2RT2RY2R^2R3R3R3R2R2R2R2R2R2R2R2S2 S2S2/S2.S23S2;S2=S2@S2BS2ASu27SIL@           @  @         @         @           @    @    @       "I    ? $?4|F̛]h$ h FXTO"dFHE#9F#9W 8 e 2 .ZE3 [E8/@?=??U MEV_F b+FF5N&*b.)N17ڼoߗ]y-l X,cΛ8x"RLH!R{  ݖl / rQՄn)G>Jeą 8-  PAl ]%-M+4cѢ6R|~kLF^if-$&N|.  l Z#7mi T̓595m>55c15-I{xY(s  l g-NNk<;wnÇt7,71Φfc2kp}3| F"%3) &o<|  PRl <G?^jg}f̮{ cP_9=  ڵl 9CwD!Ԅp>mY6N4/>U8㆜^0   l QR_@A8BC9DEF&11?++888888888888888889 99"9/9<9I9V9c9m9z9999999999999999@              "4    ? $?4|F̛] h FXTO"dFHE#9F#9W 8 e 2 .ZE3 [E8m@?fB??UQE+F n<ϧ^1WS8ժpnwK\bHx>PU諫g?<ѬȘ47W^O EĮW)LL\$D  ޼HmX##v u  @A8B]C,:DEF811?FF[9[9[9\9\9\9\9Y9V9R9F999-9#9"8888888888888999$959D9V9[9y9}999999999999}9z9y9 ~99*90989?9?9?9: :,:':+:+:+:$::: ::@                       "v    ? $F?4|F̛]2 h FXTO"dFHE#9F#9W 8 e 2 .ZE3 [E8m@?fB??UQE+F c)Rt1NsX°b9BQÚG>*GEue1K>fXуx1c՝a,ɏD,&&<ۿdurӿ}6+qͦ/ jUZT;G?Ve,Rŋ,%KU?OP  =]Im\## v  >@A9BC[:DEF&11?++19190 90 909,!9&,9!59A9K9 [9l9}999999::2:@:N:X:'Z:E]:dR:y>:3:"::::s:e:T:J:?':*=:3N:6T:=T:?V:@              "C    ? $?4|F̛] h FXTO"dFHE#9F#9W 8 e 2 .ZE3 [E8m@?fB??UQE+F }=g5sN-ngX<#Z/]w3W`^\.SMU"In|ow#ULg<0.nLDbQ)%H  Jmg#$ w  @V ABk"CDEF`11?ssb n e j j n o o o o o k k m q k h g b _ W V T [ ] ] ] ] Y V V V Y ` c e h             ! , : F T a o |  !!! ! ! !!!!!! ! ! ! ! !!!!&!/!9!C!M!X!d!o!y!!!!!E"E"E"E"L"K"L"M"L"L"K"Q"Q"/Q"AP"RU"cY"pX"]"a"d"g"j"h"h"-0@            @          @       @        "'    ? $?4|F̛]* h FXTO"dFHE#9F#9W 8 e 2 .ZE3 [E8m@?=??UQEUU)E h3"&"Vd1o]x8qǁ@ș]oW[Vqujos3L&Ģf%1(s  5Rsm V-<>w75\Ses<AhE^5].1u02\$L߃  sm N%z%qzfS8\||N\Ҧ.SLJ|בyZfIŦ&.s7  Fktm O'*'+ή&&a˜.<8\u@xF3q4yּ=<876579?!F%M)U+].e1n2v346430+&$ &&&$%$&'(((*+,-.0369;=>@"E(L0N&Q#S"V X {{yvsoie``_^ejqz (,47 ;;=@A@@?:5*#/Q`n~{TX@@                       @          @                      "    ? $Z?4|F̛]F: h FXTO"dFHE#9F#9W 8 e 2 .ZE3 [E8m@?fB??UQEUU)E B0a&pb9/LpÈ`=bZΩ)@  0VSm ihgٮ 3D^ql𹨱SSQYέx2M+vL̮ۿv@FIQXcnx~ysnhb _!]5WIW]WoVYZYZQQRSSTTSQQQTV[\^`a``\1^6b@b8g8uwtvxvw{}|m_WPHGEDEJJOTUWYYYVU U&V*W0[0`0e/p-z)*+1<I\qhI* ;<@      @               @        @                      "l    ? $<?4|F̛](R& h FXTO"dFHE#9F#9W 8 e 2 .ZE3 [E8m@?=??UQEUU)E C- ZgVzdgU/qq(5l71Ƿ  PϞm u@&'3cn3\u7ruqlT8q鐃&&"jjbb0^.&!5P_Mo^3-ܢbI)~  Cm O!1"c88s ߁װ&)/DEOp:q113rD"S)  0b͟m X!!bsYӅ1<1Zќ*j#Wf[ǞShLUq5uh5p[f퐪jXguszh fE1).SLLHI1&&g  Pm6 !YX V h+ Y3  s"*?t Z s *X99jJ?"`V h+s  [  t@#&A*:B/C<DEF11?AA)&;P*(;T*;S*;N*:G*:=*:-*:*:*:):):):):):):):)[;)Y;)Z;)X;)Z;)Z;)\;)f;)q;)|;); *;*;*;$*;/*<6*!<8*6,b;9,];5,X;0,T;*,Q;#,M;,M;,M;,M;+O;+U;+Z;+a;+k;+w;+;+;+;+;+;+;+;+;+;",;9,;H,;Y,;b,;f,;i,{;i,q;i,p;h,p;h,o;d,v;a,~;_,;\,;[,;[,;[,;Y,;_,;b,;d,;g,;-Y:$-z:$-:%-:&-:(-:(- ;(-;;%-U;%-p;%-;)-;)-;)-;(-;(-;-D;-U;-d;-t;-;-;-;-;-;-;.:-:-:-:-:-:-:.*:.@:.V:.m:.:.:.:. ;./;.V;.};.;.;.;.;.<.<.<.<_.;k.;c.;q.;.;.;.;.;.;.;'/S;$/[;'/];+/d;2/o;8/{;A/;J/;S/;`/;c/;g/;j/;/E;/k;/;/;/;/ >5]x\*b1.89DsԃƅQ#3w7-xF\5ۃቛDҐ( Jmϯ  ` l v>BhBw) 5zFq9-4B?ߠ_bjU5B'5<+Xֵ1y4u)&  11$LL'>1  *׀l ==?nlra<zY*q Ɣ޵o[  @@l 0 FFoSyi[dUՅa!36a  ol *7Ǒe&c>gV_|L  匁l EHVHOÎE5 א(HYB#|4Þ#iθt  02l 6 (X#=zG<>Pd//Ŝ@  l 6 (7e א>{ei6Ȅd'^4@  R l 6KHKU'Hd>ufUx[a"e8  +lY \  H@0AX:B6Cf<DE0F11?71J;31F;51D;.1B;)1@;!1B;1D;1H;1T;0^;0q;0;0;0;0;0;1<1<&1;51;:1;;1;=1;61{;,1_;'1O;1C;16; 1*;15;10;2:2:2:2:2:~2:v2v:e2c:L2Y:32X: 2X:2^:1i:1u:1:1:1:1:1;1&;1H;1i;1;1;1;1;1;1;1;1;1;1;1;1;1;1;1;2;2;!2;*2;o3:p3:r3:x3:3:3;39;3_;3;3;3;3;}3;v354<<5,zYqF Bҥ1J.<޵ZT7wdDI+V  ~l V%S S 6f3UFc>cm22!K0JH)N4ӂ<)$L  6ɂl : A<$Q 8<g%3ֆ@AZ@  Kl k4U>FW\Zczsk,  l *7_d`uy  l f-\r\U|ov7b 4ILwh1ڰ| %`%x-H؍eLx\W5(S.'>  {l 6 hs\4x jjl _'}r>bӴ s{ 8RZf݋TvgBG Sm*Z#Ys~ Ѓ9LEEDTsv  -^llJ  ]  @A ;B-CIDE Fn11?i#0;8#&;%# ;";v";!J;!t; ;j<%??cY@ AAoBCDKEFiG9NHH*$IkI!I"J#I%I%I&lH6'H'G'G(EY)IDk)Br)BV)kA)@(\@(?j(d?'Y>'==&<&<%i<%E<%;$;V#x;W#y;Y#y;Z#z;+<+<+<+<+<}+<{+*C>+a>&+v>=+>V+>o+>,2=,4=,2=,;=,J=!,X=(,f=5,=D,=T,=d,=r,=,=,=,=,=,S=,y=,=,=,=,>,G>,d>,>,>,>,>,>-<-=-?=-b=-=-=- >->>-r>->->->->- ?- ?- ?48@                   @         @     @    @     "    ? $Ǽ?4|F̛] h FXTO"dFHE#9F#9W 8 e 2 .ZE3 [E8/@?J??U ME_F **7WkZvbqxYN"kuqi^3mo3QXTLTfjڸ&&!xMLc-|Sđ5jLb3F*+sLD³W:fpL'YŲfMb7R7θe9Quh"Qy 8Ƴn*&Q &~6~sPD.<]PbuxL]jtb)¦Bpg(EgJ79b.g1&j`\Rbc3s;]oeu:65gMgULMB%r ^6'fUFu;TԥY\bzGb.i[aPN/t/Yֲ*, ae%, ,,,E RE @YbJ,,X6T,K,ܠYIDRʊJK*XYeBT\d6%.p  pkPul _)@h@o\q̌pƼ vc? ÄپFarŏE_D٫ԥ!)]8rރy'3110& _??  PxIl <AAQS};/|kRm¤ӝg~  ƅl 7DPDag^| F>͘hcvЄ`5;o  %l CNv< zXȰApMu`Ë[ 8c9)c4  `p6ll  D% ^  @AMB!CRDEF11?MMMMMMMMMMMMMMMMMMNwNkN^%NR0ND:N8FN-RN#]NjNvNNNNNNNNNNNOOOO O O!O%O O$OOOOOOOOO{OrOfO]OQO=O.O OOOOPP &P4P!BP.LPE]P\lPoPPPPPQ)Q>QSQeQpQQQyQiQ[QD|Q/oQaQQQBQ2Q&Q(Q$Q$QQ QQQQQQQQQQ!Qv&Ql*Qb1QS;Q=JQ8]Q5hQ8sQ@yQPQ[QnQQQQQyQmQ\QIQ7Q!Q QPPPPPPP>Q9Q7Q5Q0Q/Q,Q*Q*Q*Q+Q-Q/Qx2Qn8Qg=Q^CQTLQOVQGdQBiQ>yQ:Q=QHQUQbQuQQQQ=]Q>hQ<lQ=yQ>Q9Q:Q:Q0Q1Q2Q4Q5QzPPPPPPP@mQCeQGeQD\QBWQ8SQ4QQ,NQ!LQMQNQNQRQVQZQ_QfQlQuQQQQQQQQQQQQQ*Q2Q<|QBrQEpQFeQElQDuQE|QGQGQKQOQUQ\Q*PNPqPPPPQ @QaQQQQQPPPPPPPPPPPP|PlP]PYQUQb/Ql@Q{ZQpQQQQQQQQtQ\QBQ,QQQQQQ_QXQaQYQXQWQWQWQWQWQVQTQTQTQRQWQ\QbQnQzRQQQQQQQQQsQpQpQpQpQ R0RIRcR}RRRR R R R R R {R\R}R;R7R2R0R-R*R,R0R2R7R>uRBSRIRNQ\QbQfQwrQiQcQhQpQQQQQQQQRR RxRmRbQXQQQQQQQQQQxQkQcRc%RcuoR]r+Us5q~722M"'HU'8qsCTcکh@`6T  -l <#G$Ұ^y_w,R4|Ć9y5@  P~l i.fX]1\'Y|=8>00071Qu58bPy.7js)! }}{  l ^)$/L35qPfriÚ<+Ƹ)z3k1)>}  ӗl q3&1qzN/Y5uUnIsus~wiY4qa;ǟ{X1C#tL7Lx3(&fa).kLDDUR&6@  0 l :,-Z3LT/ß8ʺqpbieҎyV8kˇ-X)l Շ&j`zַſVHTD'32ɅBD_?L  ˍl c*!2!-utpkT9sU33<'F15n7w< ["3e"bPD0~  l S""χz㋋&1U8yhO69^5s '+Kבshe3)?  @el :JU~uŷ&dyd*mÄ:ֳ×  l 9HdvZXK h9&(mSC(SD(SD(SE(SG(S?(S!)dS )^S )WS)QS)LS)AS)>S ):S)3S(1S(.S(2S(3S(5S(,S8,S1,S+,S",S,S,S,S,S+S+"S+-S+8S+FS+SS+cS+mS+wS ,qS,mS&,gS0,ZS=,PSA,MSD,JSH,GSG,MSG,TSF,ZSD,cSD,nSI,uSN,xSP,ySR,tS,7R,[R,R,R,R,R, S,"S,;S,TS,aS,mS,zS@           @             @     @  @   @           @                @          @    @           @    @   @             @    "'    ? $?4|F̛] > h FXTO"dFHE#9F#9W 8 e 2 .ZE3 [E8/@?=??U MEV_F ]2;p;q L&"U5.1H1E]\L"`UJSSzn<2Kfn20D@LLK>H  `|l t7h. tWaaVFtc~|qF`ǣ]ŇG55 ˎ3\Mx^'ƫSk8s;ϲvҨYRK*,~y  @l <blXtJZ[-88|7qRw@  PKl . sO"oF9Rx  Lxl 3 Xڰ8hX+mۥC6c#  Wl \(6h6k&&.Tqqo0εZiS5pW:%kWfW+&B'H  Гl }A: :fq)LEF1QUK-32M]@848KY~y 1]yWUZcZq〃8wmq10&$S3q3uJ̀  pl X(;\;W狫'QEo1:˜\887|u8֜S9^<~^PW3$!DH  `Pl 9==}UH?779\f:W>qyuzV\9jk^(7~7]¯WNg6yx9q)  0R0R0R0R+0R90RI0S[0Sj0So0%Sp0'So0,S]01SL00S:00S00S0+S/(S/&S/(S/#S0S0S1R1R1R 1R,1R81RC1RK1RS1R\1R_1Rb1Rd1R_1R\1RP1RH1R71R)1R1R1R0R0R0R0S0(S07S 1GS1GS,1FSA1;SQ14S2Q2Q 2Q"2R$2FR$2rR#2R"2R2R2R2S2S2S1R1R1R1R1R1R2R&2R<2RO2RT2RY2R^2R3R3R3R2R2R2R2R2R2R2R2S2 S2S2/S2.S23S2;S2=S2@S2BS2ASu27SIL@           @  @         @         @           @    @    @       "I    ? $?4|F̛]h$ h FXTO"dFHE#9F#9W 8 e 2 .ZE3 [E8/@?=??U MEV_F b+FF5N&*b.)N17ڼoߗ]y-l X,cΛ8x"RLH!R{  ݖl / rQՄn)G>Jeą 8-  PAl ]%-M+4cѢ6R|~kLF^if-$&N|.  l Z#7mi T̓595m>55c15-I{xY(s  l g-NNk<;wnÇt7,71Φfc2kp}3| F"%3) &o<|  PRl <G?^jg}f̮{ cP_9=  ڵl 9CwD!Ԅp>mY6N4/>U8㆜^0   l QR_@A8BC9DEF&11?++888888888888888889 99"9/9<9I9V9c9m9z9999999999999999@              "4    ? $?4|F̛] h FXTO"dFHE#9F#9W 8 e 2 .ZE3 [E8m@?fB??UQE+F n<ϧ^1WS8ժpnwK\bHx>PU諫g?<ѬȘ47W^O EĮW)LL\$D  ޼HmX##v e  @A8B]C,:DEF811?FF[9[9[9\9\9\9\9Y9V9R9F999-9#9"8888888888888999$959D9V9[9y9}999999999999}9z9y9 ~99*90989?9?9?9: :,:':+:+:+:$::: ::@                       "v    ? $F?4|F̛]2 h FXTO"dFHE#9F#9W 8 e 2 .ZE3 [E8m@?fB??UQE+F c)Rt1NsX°b9BQÚG>*GEue1K>fXуx1c՝a,ɏD,&&<ۿdurӿ}6+qͦ/ jUZT;G?Ve,Rŋ,%KU?OP  =]Im\## f  >@A9BC[:DEF&11?++19190 90 909,!9&,9!59A9K9 [9l9}999999::2:@:N:X:'Z:E]:dR:y>:3:"::::s:e:T:J:?':*=:3N:6T:=T:?V:@              "C    ? $?4|F̛] h FXTO"dFHE#9F#9W 8 e 2 .ZE3 [E8m@?fB??UQE+F }=g5sN-ngX<#Z/]w3W`^\.SMU"In|ow#ULg<0.nLDbQ)%H  Jmg#$ h  @V ABk"CDEF`11?ssb n e j j n o o o o o k k m q k h g b _ W V T [ ] ] ] ] Y V V V Y ` c e h             ! , : F T a o |  !!! ! ! !!!!!! ! ! ! ! !!!!&!/!9!C!M!X!d!o!y!!!!!E"E"E"E"L"K"L"M"L"L"K"Q"Q"/Q"AP"RU"cY"pX"]"a"d"g"j"h"h"-0@            @          @       @        "'    ? $?4|F̛]* h FXTO"dFHE#9F#9W 8 e 2 .ZE3 [E8m@?=??UQEUU)E h3"&"Vd1o]x8qǁ@ș]oW[Vqujos3L&Ģf%1(s  5Rsm V-<>w75\Ses<AhE^5].1u02\$L߃  sm N%z%qzfS8\||N\Ҧ.SLJ|בyZfIŦ&.s7  Fktm O'*'+ή&&a˜.<8\u@xF3q4yּ;P*(;T*;S*;N*:G*:=*:-*:*:*:):):):):):):):)[;)Y;)Z;)X;)Z;)Z;)\;)f;)q;)|;); *;*;*;$*;/*<6*!<8*6,b;9,];5,X;0,T;*,Q;#,M;,M;,M;,M;+O;+U;+Z;+a;+k;+w;+;+;+;+;+;+;+;+;+;",;9,;H,;Y,;b,;f,;i,{;i,q;i,p;h,p;h,o;d,v;a,~;_,;\,;[,;[,;[,;Y,;_,;b,;d,;g,;-Y:$-z:$-:%-:&-:(-:(- ;(-;;%-U;%-p;%-;)-;)-;)-;(-;(-;-D;-U;-d;-t;-;-;-;-;-;-;.:-:-:-:-:-:-:.*:.@:.V:.m:.:.:.:. ;./;.V;.};.;.;.;.;.<.<.<.<_.;k.;c.;q.;.;.;.;.;.;.;'/S;$/[;'/];+/d;2/o;8/{;A/;J/;S/;`/;c/;g/;j/;/E;/k;/;/;/;/ >5]x\*b1.89DsԃƅQ#3w7-xF\5ۃቛDҐ( Jmϯ  ` l v>BhBw) 5zFq9-4B?ߠ_bjU5B'5<+Xֵ1y4u)&  11$LL'>1  *׀l ==?nlra<zY*q Ɣ޵o[  @@l 0 FFoSyi[dUՅa!36a  ol *7Ǒe&c>gV_|L  匁l EHVHOÎE5 א(HYB#|4Þ#iθt  02l 6 (X#=zG<>Pd//Ŝ@  l 6 (7e א>{ei6Ȅd'^4@  R l 6KHKU'Hd>ufUx[a"e8  +lY K  H@0AX:B6Cf<DE0F11?71J;31F;51D;.1B;)1@;!1B;1D;1H;1T;0^;0q;0;0;0;0;0;1<1<&1;51;:1;;1;=1;61{;,1_;'1O;1C;16; 1*;15;10;2:2:2:2:2:~2:v2v:e2c:L2Y:32X: 2X:2^:1i:1u:1:1:1:1:1;1&;1H;1i;1;1;1;1;1;1;1;1;1;1;1;1;1;1;1;2;2;!2;*2;o3:p3:r3:x3:3:3;39;3_;3;3;3;3;}3;v354<<5,zYqF Bҥ1J.<޵ZT7wdDI+V  ~l V%S S 6f3UFc>cm22!K0JH)N4ӂ<)$L  6ɂl : A<$Q 8<g%3ֆ@AZ@  Kl k4U>FW\Zczsk,  l *7_d`uy  l f-\r\U|ov7b 4ILwh1ڰ| %`%x-H؍eLx\W5(S.'>  {l 6 hs\4x jjl _'}r>bӴ s{ 8RZf݋TvgBG Sm*Z#Ys~ Ѓ9LEEDTsv  -^llJ  L  @A ;B-CIDE Fn11?i#0;8#&;%# ;";v";!J;!t; ;j<%??cY@ AAoBCDKEFiG9NHH*$IkI!I"J#I%I%I&lH6'H'G'G(EY)IDk)Br)BV)kA)@(\@(?j(d?'Y>'==&<&<%i<%E<%;$;V#x;W#y;Y#y;Z#z;+<+<+<+<+<}+<{+*C>+a>&+v>=+>V+>o+>,2=,4=,2=,;=,J=!,X=(,f=5,=D,=T,=d,=r,=,=,=,=,=,S=,y=,=,=,=,>,G>,d>,>,>,>,>,>-<-=-?=-b=-=-=- >->>-r>->->->->- ?- ?- ?48@                   @         @     @    @     "    ? $Ǽ?4|F̛] h FXTO"dFHE#9F#9W 8 e 2 .ZE3 [E8/@?J??U ME_F **7WkZvbqxYN"kuqi^3mo3QXTLTfjڸ&&!xMLc-|Sđ5jLb3F*+sLD³W:fpL'YŲfMb7R7θe9Quh"Qy 8Ƴn*&Q &~6~sPD.<]PbuxL]jtb)¦Bpg(EgJ79b.g1&j`\Rbc3s;]oeu:65gMgULMB%r ^6'fUFu;TԥY\bzGb.i[aPN/t/Yֲ*, ae%, ,,,E RE @YbJ,,X6T,K,ܠYIDRʊJK*XYeBT\d6%.p  pkPul _)@h@o\q̌pƼ vc? ÄپFarŏE_D٫ԥ!)]8rރy'3110& _??  PxIl <AAQS};/|kRm¤ӝg~  ƅl 7DPDag^| F>͘hcvЄ`5;o  %l CNv< zXȰApMu`Ë[ 8c9)c4  `p6ll  D% M  @AMB!CRDEF11?MMMMMMMMMMMMMMMMMMNwNkN^%NR0ND:N8FN-RN#]NjNvNNNNNNNNNNNOOOO O O!O%O O$OOOOOOOOO{OrOfO]OQO=O.O OOOOPP &P4P!BP.LPE]P\lPoPPPPPQ)Q>QSQeQpQQQyQiQ[QD|Q/oQaQQQBQ2Q&Q(Q$Q$QQ QQQQQQQQQQ!Qv&Ql*Qb1QS;Q=JQ8]Q5hQ8sQ@yQPQ[QnQQQQQyQmQ\QIQ7Q!Q QPPPPPPP>Q9Q7Q5Q0Q/Q,Q*Q*Q*Q+Q-Q/Qx2Qn8Qg=Q^CQTLQOVQGdQBiQ>yQ:Q=QHQUQbQuQQQQ=]Q>hQ<lQ=yQ>Q9Q:Q:Q0Q1Q2Q4Q5QzPPPPPPP@mQCeQGeQD\QBWQ8SQ4QQ,NQ!LQMQNQNQRQVQZQ_QfQlQuQQQQQQQQQQQQQ*Q2Q<|QBrQEpQFeQElQDuQE|QGQGQKQOQUQ\Q*PNPqPPPPQ @QaQQQQQPPPPPPPPPPPP|PlP]PYQUQb/Ql@Q{ZQpQQQQQQQQtQ\QBQ,QQQQQQ_QXQaQYQXQWQWQWQWQWQVQTQTQTQRQWQ\QbQnQzRQQQQQQQQQsQpQpQpQpQ R0RIRcR}RRRR R R R R R {R\R}R;R7R2R0R-R*R,R0R2R7R>uRBSRIRNQ\QbQfQwrQiQcQhQpQQQQQQQQRR RxRmRbQXQQQQQQQQQQxQkQcRc%RcuoR]r+Us5q~722M"'HU'8qsCTcکh@`6T  -l <#G$Ұ^y_w,R4|Ć9y5@  P~l i.fX]1\'Y|=8>00071Qu58bPy.7js)! }}{  l ^)$/L35qPfriÚ<+Ƹ)z3k1)>}  ӗl q3&1qzN/Y5uUnIsus~wiY4qa;ǟ{X1C#tL7Lx3(&fa).kLDDUR&6@  0 l :,-Z3LT/ß8ʺqpbieҎyV8kˇ-X)l Շ&j`zַſVHTD'32ɅBD_?L  ˍl c*!2!-utpkT9sU33<'F15n7w< ["3e"bPD0~  l S""χz㋋&1U8yhO69^5s '+Kבshe3)?  @el :JU~uŷ&dyd*mÄ:ֳ×  l 9HdvZXK h9&(mSC(SD(SD(SE(SG(S?(S!)dS )^S )WS)QS)LS)AS)>S ):S)3S(1S(.S(2S(3S(5S(,S8,S1,S+,S",S,S,S,S,S+S+"S+-S+8S+FS+SS+cS+mS+wS ,qS,mS&,gS0,ZS=,PSA,MSD,JSH,GSG,MSG,TSF,ZSD,cSD,nSI,uSN,xSP,ySR,tS,7R,[R,R,R,R,R, S,"S,;S,TS,aS,mS,zS@           @             @     @  @   @           @                @          @    @           @    @   @             @    "'    ? $?4|F̛] > h FXTO"dFHE#9F#9W 8 e 2 .ZE3 [E8/@?=??U MEV_F ]2;p;q L&"U5.1H1E]\L"`UJSSzn<2Kfn20D@LLK>H  `|l t7h. tWaaVFtc~|qF`ǣ]ŇG55 ˎ3\Mx^'ƫSk8s;ϲvҨYRK*,~y  @l <blXtJZ[-88|7qRw@  PKl . sO"oF9Rx  Lxl 3 Xڰ8hX+mۥC6c#  Wl \(6h6k&&.Tqqo0εZiS5pW:%kWfW+&B'H  Гl }A: :fq)LEF1QUK-32M]@848KY~y 1]yWUZcZq〃8wmq10&$S3q3uJ̀  pl X(;\;W狫'QEo1:˜\887|u8֜S9^<~^PW3$!DH  `Pl 9==}UH?779\f:W>qyuzV\9jk^(7~7]¯WNg6yx9q)  0R0R0R0R+0R90RI0S[0Sj0So0%Sp0'So0,S]01SL00S:00S00S0+S/(S/&S/(S/#S0S0S1R1R1R 1R,1R81RC1RK1RS1R\1R_1Rb1Rd1R_1R\1RP1RH1R71R)1R1R1R0R0R0R0S0(S07S 1GS1GS,1FSA1;SQ14S2Q2Q 2Q"2R$2FR$2rR#2R"2R2R2R2S2S2S1R1R1R1R1R1R2R&2R<2RO2RT2RY2R^2R3R3R3R2R2R2R2R2R2R2R2S2 S2S2/S2.S23S2;S2=S2@S2BS2ASu27SIL@           @  @         @         @           @    @    @       "I    ? $?4|F̛]h$ h FXTO"dFHE#9F#9W 8 e 2 .ZE3 [E8/@?=??U MEV_F b+FF5N&*b.)N17ڼoߗ]y-l X,cΛ8x"RLH!R{  ݖl / rQՄn)G>Jeą 8-  PAl ]%-M+4cѢ6R|~kLF^if-$&N|.  l Z#7mi T̓595m>55c15-I{xY(s  l g-NNk<;wnÇt7,71Φfc2kp}3| F"%3) &o<|  PRl <G?^jg}f̮{ cP_9=  ڵl 9CwD!Ԅp>mY6N4/>U8㆜^0   l QR_PU諫g?<ѬȘ47W^O EĮW)LL\$D  ޼HmX##p V  @A8B]C,:DEF811?FF[9[9[9\9\9\9\9Y9V9R9F999-9#9"8888888888888999$959D9V9[9y9}999999999999}9z9y9 ~99*90989?9?9?9: :,:':+:+:+:$::: ::@                       "v    ? $F?4|F̛]2 h FXTO"dFHE#9F#9W 8 e 2 .ZE3 [E8m@?fB??UQE+F c)Rt1NsX°b9BQÚG>*GEue1K>fXуx1c՝a,ɏD,&&<ۿdurӿ}6+qͦ/ jUZT;G?Ve,Rŋ,%KU?OP  =]Im\## W  8@A9BC[:DEF&11?++19190 90 909,!9&,9!59A9K9 [9l9}999999::2:@:N:X:'Z:E]:dR:y>:3:"::::s:e:T:J:?':*=:3N:6T:=T:?V:@              "C    ? $?4|F̛] h FXTO"dFHE#9F#9W 8 e 2 .ZE3 [E8m@?fB??UQE+F }=g5sN-ngX<#Z/]w3W`^\.SMU"In|ow#ULg<0.nLDbQ)%H  Jmg#$"Z ^ ( 3 3  s"*?n  c $X99?"`^ ( 3   v@AB,CCKDEF.1jJ?77-- - -(1%!W  7/!Mi#!%<'\+1/6$80<>@LDCEFjGH;"5J$[K'KR+L/K2J5J[8fH{:FE@BC=: C2B.A*\@:&?#=: |;;:8t-7/W5t3iA1*/-*@                  "6    ? $?4|F̛] h FXTO"dFHE5F5W 8 e 2 .ZE3 [E80?^?9vQE:)E #(##JnmkZeqx\Iq7RiDiKxMb*QfL_)%UU/0 "&2TEF:]+WdB&*TDLF#EE\MD kTDE&9Su5x}1Q\LNcWm1MaM39VnyJfɭw7n&w=7vg3٪D\ULDFى1T5WLȦ+UJDDE+Dj*/iX1QÞ9ffʲ-RTX%( DYbRŕ,(.l,.T%%\J`%@J&``XX *YQeYc`\T@&ɰ%.YA(,-  cWj #*1   @ AMBCRDEPFH1jJ?TT\lM^hMi]Mi]MiWMe\MhVMjQMg\MiVMfYMoOMlRMiTMdWM`]MRqMIM=M M N>NNiN-MOOsaP7P%$Q\Q Q QQQo0Mt&Mv#MoMz(M5MDM MxxN OO5O`HPyPPP QQQ3P-P PPPOOOOOOOOOOPDPPP QRQQQQ%RCR`R}RR RR!$@           @      @          "    ? $ǹ?4|F̛]Rn h FXTO"dFHE5F5W 8 e 2 .ZE3 [E8\?^?9vQE8)E p15ק= Y]n3s|c81h\/D3S4pU8kVvfP] 6$sG ^&\CrJDEF1jJ?AA4ZJ4TJ4QJ4RJ4VJ4LJ4WJ4SJ4RJ4UJ4WJ4XJ4]J4^J4lJ4lJ4oJ4pJ4nJ4iJ4kJ4fJ4dJ4cJ4aJ4ZJ4\J4ZJ4PJ4`J4XJ4XJ4YJ4TJ4QJ4MJ4JJ6H6H6H6H6H6H6H6H6H6H6H6H6|H6zH6uH6sH6pH6sH6wH6}H6H6Hs6H`6HN6H!6H5H5I55I5NI5hI5}I5I5I5I5I5I5I5I5J5J5#J6'J96-J`6+J6&J6!J6J6 J6I7I7I07I;7ID7IM7lIR7TIN77IF7H!7H6H6H6H6sH6oH6YH6VH6XH6YH6[H8XH8OH8IH8?H87H86H8*H8H8H8H8H8Hg8HK8 H&8H 8H7$H7>H7YH7|H7H8H8H'8I28,ID8dIO8IB8I=8I18J8"J70J7J7J7I7I7I7FI7I7H7H8H78Hb8zH8_H8LH8:H9#HA9H_9Hz9G9G:LD:SD:MD:OD:YD:MD:TD:WD:[D:`D:mD:xD:D:D:D:D:D:D:D:D:D:D:D:D:D:D:D:D:D:D:D:D:D:DtBG>BF>BG>BI>BP>BM>BK>B@>B=>B5>B->B$>B>C>+C>>C=`C=C=C=C=Cz=!Dc=MDP=DI=DC=D\=Dy=E='E=8E=>E=CE >AE%>4EJ>"Eg> E{>D>D>D>D>D}>Dk>DM>D)>D>D=D=D=D=E=E=1E=AE=CE=JEux@            @                       @                 @           @                        @                  "    ? $?4|F̛] h FXTO"dFHE5F5W 8 e 2 .ZE3 [E8\?^?9vQE8)E d+\U콰KҲC.jO eѯXU2U;7˯ ԦW50LE0< g9)LLH>Ouz ?@"me|@8 c1n1ygj.%rbk8R/Wk7sYNs9'W|VRR嚕]Jjqu1H Rb2#pkf*b' s;벀G8L$Ș&&"@D`"IM] ?@# Ap|@8 K44u-ɫִbi[3{oh^]8GY1msȃQRܻ㋊Su*tDJǟ}sq1-En`󼈇e32DJ$"]@  ؒ.W ^-77sߦj&5uQ* נKxKuޫ.|U V,_m`oaQEB1E ` ?@"Ձ|@8 b:R:C:xsqJ3x9VqgsSF+1Yg7zuWII-njjUq0TLU1wmeisUSS>PxI 0b@"DIDĢE{5t  ڎ1W M=8=5tc\f.1,jާ5fs7nۋp,jHH3" EF#]c,qea0UcU  $2Wh@& *   &@."A"BB&CDDE$F1jJ?_"DY"D["DY"DV"DU"DT"~DS"|DQ"wDQ"wDQ"vDP"oDP"oDU"iDM"rDO"hDP"dDR"fDN"eDK"dDJ"_DG"`DD"aDB"bDD"`DH"]DE"YDD"TDC"ODA"ID@"CD@"AD@"AD?"@D>"ID<"WD:"aD6"qD2"D."DY$BX$BW$BV$BW$BW$B[$B_$zB]$|Ba$sBc$nBd$iBf$cBg$cBg$cBh$cBh$eBf$nBe$oB`$mB_$mBa$tB]$qBc$tB_$rB[$pB\$oBY$nBX$nBV$lBU$lBS$kBR$jBO$jBL$jBI$iBE$jBB$kB<$oB8$qB2$tB,$xB&${B$B$B $B#B#B#B#B#B#B#B#B#Cq#Ce#"CX#3CG#IC6#fC/#C*#C&#C)#C,#C;#CG#C[# Dz#D#)D#:D#AD$6D!$.D:$#DP$Di$D$C$C$C$C$C$C$C$lC$\C$IC$9C$,C$C$C$B$B$B$Bq$Ba$BW$BM$vBB$hB:$^B4$TB+$TB+$XB*$YB-$[B&OB&MB!&LB&UCmˈ-6!PJP% D,,d  ٷ7W R<>Y86wTrG' FncsrjF8]umHFHU<{߷>]zJwIbiC<\9[N*5<Ѓ fyiQ!"%&&&$LJbbh@?P ?@"%|@8%3'q   @y(AFB,CIDE(Fl1jJ?(TH(UH(ZH(VH(QH(OH(SH(TH(XH(XH(YH(YH(YH}(YH(YH(YH(YH~(YH(YH(YH(ZH(YH}(XH{(XHy(WH~(WH~(VH(SH(QH(PH(NHD*JGA*AG>*>G<*4G;*-G>* G9*G8*G1*G/*G+* G *G*G *G)G)!G))G)1G)H**H* H{*Gw*Gm*G`*GT*GH*|G;*hG3*[G-*SG!*JG*TG*LG *YG$*fG)*hG1*rG+F*F+F*F*F*F*F*F*F+F +F+F!+F6+FJ+F_+Fx+F+F+F+G+G+G+G+/G,AG ,aG,G+G+G+H+-H+VH+H+H+H+H+H+H+ I+I+ I+ I+ I34@          @                     @              "-    ? $?4|F̛]J0 h FXTO"dFHE5F5W 8 e 2 .ZE3 [E8\?^?9vQE8)E V*oSԐf/yDDr7P7Ɣuь`B0.r  ,W _R[WnQq!(LiUDR&migIPFPS>WǷ^uu1Q4QK\*Gyg371-125oMMV.onfe$LJ."IL&$ELD0L%)>P  ],W n5~dž|"D79r1SLOOLJntD֢88bfg;33؃K%\I9߫h  ,WJ(NY)z   @..AmIB41C6KDEF1jJ?1.K1. K1. K/.K/. K/.K1.K8.K9.K9.K9.'K<.#K@.*KC.0K;.2KE.4KH.5KI.3KL.2KU.0KY.-K`.*Kc.)Ke.)Kh.(Ki.&Kf. Kh.Kj.Kk.Km.Ka.Ki.Kn.Kw. Ka.Kx. K{.J.J.J.J.J.J/I/I/I/I/I/I/I/I/I/I/I/I/I/I/I{/It/Io/Ij/Id/I]/IV/IC/J;/6J,/TJ /mJ/J/J/J/J/J/J/J /J/J/J'/J5/KG/KW/Km/K/J/J/J/J/J/J/J/J0J0J0J0J 0J0{J/mJ/bJ/WJ/GJ/:J/,J/!J/J/ J/J/I/I/I/I/I/I/I/I-0{I+0{I*0zI(0~I)0{I)0xI*0vI.0pI*0rI00oI30nI:0pI@0qIJ0sIR0vI\0yIi0}It0~I0I0I0I0I0I0I0I0I0I0I0I0I0I0J0J0 J0'J0-J04J07J09J0;J0@J0EJ0JJ0WJ 1rJ1J)1J/1J21J51J01J%1J1K 1K0%K0-K0/K0-K0*K0%K}0Kn0K`0KP0 KI0KI0KE0KBD@              @                        @                      "d    ? $4?4|F̛] @( h FXTO"dFHE5F5W 8 e 2 .ZE3 [E8\?^?9vQE8)E f7##qʬ^뎧^87VDV=ѝ^-UEj면Ѓ;KeRY2s7%gmkf{[˗@Hqn ٕ$!$f6kDl5L55 3_˝5|w\$LĄJ%"H$LLJ%ĠL&I9߯  `d)W ]&&wLM^3'frQB$]kNb^pG<|no`Ȭ)gsey1q%ٯ-ogZ"6P(fpNÎ̄C5eXV1jG  rW gyfHDTpTqm1gQHf&39rn8nA*jj11WQB D\ffctF7U U"์tj36@%)%  D ~0  *WN0?82    f@CALBqGCQDEFV1jJ?iiFMFMFMFMFMFMFMF}MFMFMFMFMFMFMEMEMEMEMEMENaEYN;ENEND)OD{O}DOTDP0DGP!DdPDPCPCPCPCP#DL'DL!DL-DLADLLDMZD3MwDrMDMDMDCNDNDND4OEO#EO0EOCE PcEPwEP~EPE PYFF*O>F+O=F'O>F&OF%OBF"OFFOLFO_FOqFOFOFNFNFOGO"GO>G*OiGLOZGOEGO6GOGOG PG+P GFPGcP0GPGPpGQfG@Q3GUQGaQFzQF}QAFQEQEQcEQ$EQD|QDuQ(,@           @       @                "    ? $?4|F̛]H` h FXTO"dFHE5F5W 8 e 2 .ZE3 [E8\?^?9vQE8)E b)IIˎqx<;7OaȄiNK2%Nij0[Ds2G>ǝS733$ifs@  GW NFFTiD^gUdGT2GX\@ĤNhj97,#  @W HJJ'o:*#\!Yu|wsQ;?qǵ8] ]R[tf0f5ˍh*;W32LH"jbbf&a4%7,ߋ  pW ++C"w- !  @V-A-B4C2DEF1jJ?l-1o-1s-1v-1x-1z-1}-1-1-1}-1-1-1-1-1}-1z-1q-1p-1i-1a-1^-1Z-1W-1W-1V-1W-1[-1^-1e-1i-1q-1v-1x-1-1-1-1-1-1-1-1j0g.r0e.m0d.p0d.t0d.u0e.x0d.y0d.0e.0d.0a.0[.0Z.0Y.0\.0\.0].0a.0d.0i.0s.0x.0.0.0.~0.h0.O0/50:/0c////////#0/c0/0~/0s/"1{/g1/1/1/1/20"2-062\0@20I20H20?2152412Y12v11111111111j11\11>1121111!1o1:1R1T1;1w1%1111110102021%21%2 1$2 1$24 .4-4.4-4-4-4-4-4-4-4.4.4.4.4.4 .4 .4.4.o4*.L45.%49.3?.35.33.v32.h33.Z33.Y33.W34.V34.V34.X3=.Z3B.c3Z.m3q.t3.3.3i/3/303O03030z30v3030303030303040F40c414Q141414L242Q4232b32222222p2s2a2g2\2c2Q2]2CD@             @                         @                       "    ? $t?4|F̛]`b h FXTO"dFHE5F5W 8 e 2 .ZE3 [E8\?`?9vQE8)E b3"Z"K˭&135iԢD2oG--Ut]]Z%I\bfȃ{UQMneq("Bb-W͉  aW v' '뫂𸚔\ĥq)Lոqqq#UU [˚BN (L(Ga5uubTe UQUUcT7w=2aXb*,G9G=skI110%bbf&$(&.`} ?@"y|@8 j..Lo]zq\̫4CQX&ng)B+bYz>9yuN˧)''혖^&dJS#Zֳ{7»kU9ǎ4WnIbDĦ$I11110! g~ߏR ?@"ƃ|@8/zX "  h@"AABgAC?\DEF,1jJ?44M@KO@HKr@Jf@JK@I@:I?H>G=G=F;Ej:FE9D`7D53C4BO2+B0B.B+B)\C@'TEU%F.$H[#?K"XMr"O"Q"T#V&%~X&1Z(Z*[S0\6R\);4YU@@TqAReA?P_AN;AnMAL@LJ?!I>G>G>`G>EG@                 "    ? $|?4|F̛]h h FXTO"dFHE5F5W 8 e 2 .ZE3 [E8\?^?9vQE8)E @t@uWXsm3\S1zM37q8*.Bn3)1[I83ġ|scXq 3c'P1XE]DUW "RU¢i*α's^4995-mnjg(H>K&4=#>?@mA^BDYGcIK2NiPRT`VW3XXMYnZYeZ]`Z 'Z!Y!Y!Y!Y@                      "    ? $Ǎ?4|F̛]y h FXTO"dFHE5F5W 8 e 2 .ZE3 [E8\?^?9vQE8)E N3WS ^֮1z]AD0 *\Fk3ugQSSu2„&!(DbWLL%qEUDHLTőiL"$I11!H1jqnun ᘔ. H+V\\S0 x<V\a|k9ťLi滔Ffsk1|-EO /]bg[Mb3r'bZaXTEEjk0JƷQ \\' ku{Ĉ \Ise7L{}v"4vZX7ɼś'fe汹yŦܽţc7ybwxn{ufMks׊*kř^[wPno6՛K;7[Eeۤv]deVlۚ/77slsjϯbx猕[s,fʖyȌd2xbAD\0@DY0@DV0@DV0BDW0@DW0=DW0DW0ADX0DDR0EDU0FDW0FDX0FDU0EDW0FDY0FDW0GDT0HDT0FDT0JDV0JDS0KDW0HDU0HDV0HDY0FDX0HDW0IDZ0KD\0LDa0LDc0LDe0LDk0IDm0JDt0LDy0RD~0JD0AD0>D0;D@                       "Q    ? $!?4|F̛]  h FXTO"dFHE5F5W 8 e 2 .ZE3 [E8\?^?9vQE8)E X&&]^+tZnz\\oV^up7117 6mNU四ɸwr77;U%E,7ś z򃈉A$̦SH&a1"$D  @ WV&r ' &  @0ACB1CDDEDFj1jJ?f13Cg1-Cc1/Ch1-Cf1)Cf1)Ck1-Ci1-Cg1,Ce1-Cc1.C_1.C^1/C\10C]12C[13CZ12CZ10CY1/CW1.CS1+CR1+CP1*CK1)CH1(CC1&C@1%C:1$C61$C/1$C+1$C%1%C1&C1(C1+C1+C 1/C13C14C09C0?C0FC0NC0WC0bC0lC0tC0}C0C0C0C0C0C0C0C0C0C0C0C0C0D0 D0D0D0)D02D0=D0GD0QD0ZD0bD0hD0oD0uD0{D0D0D1D1D1D+1D71D?1DK1DR1DY1D^1Dd1}Dn1wDh1uDn1nDt1fD1`D1YD1RD1ID1@D19D11D1(D1#D1D1D1D1D1D1C1C1C1C1C1C1C1C1C1C1C1C1C1C1C1|C1oC1fC1\C1TC1JC1BC1:Cw12Cp1*Cn1#Ch1Cf1Cc1Ca1C`1C`1C_1C^1CX1C[1C]1Ca1Cc1C24@                                                "    ? $Nj?4|F̛]w$ h FXTO"dFHE5F5W 8 e 2 .ZE3 [E8\?^?9vQE8)E #}|87;P*(;T*;S*;N*:G*:=*:-*:*:*:):):):):):):):)[;)Y;)Z;)X;)Z;)Z;)\;)f;)q;)|;); *;*;*;$*;/*<6*!<8*6,b;9,];5,X;0,T;*,Q;#,M;,M;,M;,M;+O;+U;+Z;+a;+k;+w;+;+;+;+;+;+;+;+;+;",;9,;H,;Y,;b,;f,;i,{;i,q;i,p;h,p;h,o;d,v;a,~;_,;\,;[,;[,;[,;Y,;_,;b,;d,;g,;-Y:$-z:$-:%-:&-:(-:(- ;(-;;%-U;%-p;%-;)-;)-;)-;(-;(-;-D;-U;-d;-t;-;-;-;-;-;-;.:-:-:-:-:-:-:.*:.@:.V:.m:.:.:.:. ;./;.V;.};.;.;.;.;.<.<.<.<_.;k.;c.;q.;.;.;.;.;.;.;'/S;$/[;'/];+/d;2/o;8/{;A/;J/;S/;`/;c/;g/;j/;/E;/k;/;/;/;/ >5]x\*b1.89DsԃƅQ#3w7-xF\5ۃቛDҐ( Jmϯ  ` l v>BhBw) 5zFq9-4B?ߠ_bjU5B'5<+Xֵ1y4u)&  11$LL'>1  *׀l ==?nlra<zY*q Ɣ޵o[  @@l 0 FFoSyi[dUՅa!36a  ol *7Ǒe&c>gV_|L  匁l EHVHOÎE5 א(HYB#|4Þ#iθt  02l 6 (X#=zG<>Pd//Ŝ@  l 6 (7e א>{ei6Ȅd'^4@  R l 6KHKU'Hd>ufUx[a"e8  +lY @  B@0AX:B6Cf<DE0F11?71J;31F;51D;.1B;)1@;!1B;1D;1H;1T;0^;0q;0;0;0;0;0;1<1<&1;51;:1;;1;=1;61{;,1_;'1O;1C;16; 1*;15;10;2:2:2:2:2:~2:v2v:e2c:L2Y:32X: 2X:2^:1i:1u:1:1:1:1:1;1&;1H;1i;1;1;1;1;1;1;1;1;1;1;1;1;1;1;1;2;2;!2;*2;o3:p3:r3:x3:3:3;39;3_;3;3;3;3;}3;v354<<5,zYqF Bҥ1J.<޵ZT7wdDI+V  ~l V%S S 6f3UFc>cm22!K0JH)N4ӂ<)$L  6ɂl : A<$Q 8<g%3ֆ@AZ@  Kl k4U>FW\Zczsk,  l *7_d`uy  l f-\r\U|ov7b 4ILwh1ڰ| %`%x-H؍eLx\W5(S.'>  {l 6 hs\4x jjl _'}r>bӴ s{ 8RZf݋TvgBG Sm*Z#Ys~ Ѓ9LEEDTsv  -^llJ  G  @A ;B-CIDE Fn11?i#0;8#&;%# ;";v";!J;!t; ;j<%??cY@ AAoBCDKEFiG9NHH*$IkI!I"J#I%I%I&lH6'H'G'G(EY)IDk)Br)BV)kA)@(\@(?j(d?'Y>'==&<&<%i<%E<%;$;V#x;W#y;Y#y;Z#z;+<+<+<+<+<}+<{+*C>+a>&+v>=+>V+>o+>,2=,4=,2=,;=,J=!,X=(,f=5,=D,=T,=d,=r,=,=,=,=,=,S=,y=,=,=,=,>,G>,d>,>,>,>,>,>-<-=-?=-b=-=-=- >->>-r>->->->->- ?- ?- ?48@                   @         @     @    @     "    ? $Ǽ?4|F̛] h FXTO"dFHE#9F#9W 8 e 2 .ZE3 [E8/@?J??U ME_F **7WkZvbqxYN"kuqi^3mo3QXTLTfjڸ&&!xMLc-|Sđ5jLb3F*+sLD³W:fpL'YŲfMb7R7θe9Quh"Qy 8Ƴn*&Q &~6~sPD.<]PbuxL]jtb)¦Bpg(EgJ79b.g1&j`\Rbc3s;]oeu:65gMgULMB%r ^6'fUFu;TԥY\bzGb.i[aPN/t/Yֲ*, ae%, ,,,E RE @YbJ,,X6T,K,ܠYIDRʊJK*XYeBT\d6%.p  pkPul _)@h@o\q̌pƼ vc? ÄپFarŏE_D٫ԥ!)]8rރy'3110& _??  PxIl <AAQS};/|kRm¤ӝg~  ƅl 7DPDag^| F>͘hcvЄ`5;o  %l CNv< zXȰApMu`Ë[ 8c9)c4  `p6ll  D% B  @AMB!CRDEF11?MMMMMMMMMMMMMMMMMMNwNkN^%NR0ND:N8FN-RN#]NjNvNNNNNNNNNNNOOOO O O!O%O O$OOOOOOOOO{OrOfO]OQO=O.O OOOOPP &P4P!BP.LPE]P\lPoPPPPPQ)Q>QSQeQpQQQyQiQ[QD|Q/oQaQQQBQ2Q&Q(Q$Q$QQ QQQQQQQQQQ!Qv&Ql*Qb1QS;Q=JQ8]Q5hQ8sQ@yQPQ[QnQQQQQyQmQ\QIQ7Q!Q QPPPPPPP>Q9Q7Q5Q0Q/Q,Q*Q*Q*Q+Q-Q/Qx2Qn8Qg=Q^CQTLQOVQGdQBiQ>yQ:Q=QHQUQbQuQQQQ=]Q>hQ<lQ=yQ>Q9Q:Q:Q0Q1Q2Q4Q5QzPPPPPPP@mQCeQGeQD\QBWQ8SQ4QQ,NQ!LQMQNQNQRQVQZQ_QfQlQuQQQQQQQQQQQQQ*Q2Q<|QBrQEpQFeQElQDuQE|QGQGQKQOQUQ\Q*PNPqPPPPQ @QaQQQQQPPPPPPPPPPPP|PlP]PYQUQb/Ql@Q{ZQpQQQQQQQQtQ\QBQ,QQQQQQ_QXQaQYQXQWQWQWQWQWQVQTQTQTQRQWQ\QbQnQzRQQQQQQQQQsQpQpQpQpQ R0RIRcR}RRRR R R R R R {R\R}R;R7R2R0R-R*R,R0R2R7R>uRBSRIRNQ\QbQfQwrQiQcQhQpQQQQQQQQRR RxRmRbQXQQQQQQQQQQxQkQcRc%RcuoR]r+Us5q~722M"'HU'8qsCTcکh@`6T  -l <#G$Ұ^y_w,R4|Ć9y5@  P~l i.fX]1\'Y|=8>00071Qu58bPy.7js)! }}{  l ^)$/L35qPfriÚ<+Ƹ)z3k1)>}  ӗl q3&1qzN/Y5uUnIsus~wiY4qa;ǟ{X1C#tL7Lx3(&fa).kLDDUR&6@  0 l :,-Z3LT/ß8ʺqpbieҎyV8kˇ-X)l Շ&j`zַſVHTD'32ɅBD_?L  ˍl c*!2!-utpkT9sU33<'F15n7w< ["3e"bPD0~  l S""χz㋋&1U8yhO69^5s '+Kבshe3)?  @el :JU~uŷ&dyd*mÄ:ֳ×  l 9HdvZXK h9&(mSC(SD(SD(SE(SG(S?(S!)dS )^S )WS)QS)LS)AS)>S ):S)3S(1S(.S(2S(3S(5S(,S8,S1,S+,S",S,S,S,S,S+S+"S+-S+8S+FS+SS+cS+mS+wS ,qS,mS&,gS0,ZS=,PSA,MSD,JSH,GSG,MSG,TSF,ZSD,cSD,nSI,uSN,xSP,ySR,tS,7R,[R,R,R,R,R, S,"S,;S,TS,aS,mS,zS@           @             @     @  @   @           @                @          @    @           @    @   @             @    "'    ? $?4|F̛] > h FXTO"dFHE#9F#9W 8 e 2 .ZE3 [E8/@?=??U MEV_F ]2;p;q L&"U5.1H1E]\L"`UJSSzn<2Kfn20D@LLK>H  `|l t7h. tWaaVFtc~|qF`ǣ]ŇG55 ˎ3\Mx^'ƫSk8s;ϲvҨYRK*,~y  @l <blXtJZ[-88|7qRw@  PKl . sO"oF9Rx  Lxl 3 Xڰ8hX+mۥC6c#  Wl \(6h6k&&.Tqqo0εZiS5pW:%kWfW+&B'H  Гl }A: :fq)LEF1QUK-32M]@848KY~y 1]yWUZcZq〃8wmq10&$S3q3uJ̀  pl X(;\;W狫'QEo1:˜\887|u8֜S9^<~^PW3$!DH  `Pl 9==}UH?779\f:W>qyuzV\9jk^(7~7]¯WNg6yx9q)  0R0R0R0R+0R90RI0S[0Sj0So0%Sp0'So0,S]01SL00S:00S00S0+S/(S/&S/(S/#S0S0S1R1R1R 1R,1R81RC1RK1RS1R\1R_1Rb1Rd1R_1R\1RP1RH1R71R)1R1R1R0R0R0R0S0(S07S 1GS1GS,1FSA1;SQ14S2Q2Q 2Q"2R$2FR$2rR#2R"2R2R2R2S2S2S1R1R1R1R1R1R2R&2R<2RO2RT2RY2R^2R3R3R3R2R2R2R2R2R2R2R2S2 S2S2/S2.S23S2;S2=S2@S2BS2ASu27SIL@           @  @         @         @           @    @    @       "I    ? $?4|F̛]h$ h FXTO"dFHE#9F#9W 8 e 2 .ZE3 [E8/@?=??U MEV_F b+FF5N&*b.)N17ڼoߗ]y-l X,cΛ8x"RLH!R{  ݖl / rQՄn)G>Jeą 8-  PAl ]%-M+4cѢ6R|~kLF^if-$&N|.  l Z#7mi T̓595m>55c15-I{xY(s  l g-NNk<;wnÇt7,71Φfc2kp}3| F"%3) &o<|  PRl <G?^jg}f̮{ cP_9=  ڵl 9CwD!Ԅp>mY6N4/>U8㆜^0   l QR_>F((%%EPP((--:*urn:schemas-microsoft-com:office:smarttagsStreet;*urn:schemas-microsoft-com:office:smarttagsaddress9*urn:schemas-microsoft-com:office:smarttagsplace=*urn:schemas-microsoft-com:office:smarttags PlaceType= *urn:schemas-microsoft-com:office:smarttags PlaceName L %'ac :@55:: CM WbALT[U#Y#%%U+_+2233 ::z<< =0===@ @~BBCC@DJDGGLLNNNNPP\\]]f__}bbj jum}mnoxxl{p{MWPmu>HIRˌ0ŕ *¬')::::::::::::::::::::::::::::::::::::::::::::::::::KK{Q|Q}Q~QQQ]]yc{c :;LMʮʮ 45*{{kløƸ޸\d-n. 88^8`OJQJo( ^`OJQJo(^`OJPJQJ^Jo(^`OJQJ^Jo(hHopp^p`OJQJo(hH@ @ ^@ `OJQJo(hH^`OJQJ^Jo(hHo^`OJQJo(hH^`OJQJo(hH^`OJQJ^Jo(hHoPP^P`OJQJo(hH-n+        fi YX\ jX>C?Ll_4v)N'`d/6J+GC> v!'@!{#$?(M)RY),L./6i/11G2B5 6W6>:~:<\<=&)>F?L @rB9C~CYoECJKUL;QSG S@T2LT~MT4U^WuX.Y6Y%Z%[s\>^ x_Y `?6bseg hH i~j3m-o p p,sKsmsh;tyyyzP}A "i<uX}RH0;lDSqhT }7I,m2%oiM/wHrsSPfNR+dkkLh$Q kES#z<-;{[rFb)W.S VG]Mi#"W{ d0k9eC=6=Yf7 _1R6@++ {Q++(BC@J@UnknownGz Times New Roman5Symbol3& z Arial5"System?5 z Courier New;Wingdings"qhlFlF%^O^O>4dCC3QHX)?S 2Multiple RegressionRadford UniversityRadford University   Oh+'0  (4 T ` l xMultiple RegressionRadford UniversityNormalRadford University2Microsoft Office Word@F#@]@?:&@?:&՜.+,0  hp  Radford UniversityO^C Multiple Regression Title  !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~      !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~Root Entry F2N:&Data ۅ1Table$WordDocument7 SummaryInformation(DocumentSummaryInformation8CompObjq  FMicrosoft Office Word Document MSWordDocWord.Document.89q