Stat404 Fall 2009 Lab 11 1. Consider the following path model based on a sample of 250 subjects (Ss): a. Although a “just identified” path model need not be “fully recursive,” all “fully recursive” path models are “just identified.” What do these two expressions mean? (In defining each expression, be sure to explain how you can tell that these expressions apply to this path model.) b. Whenever a path model is fully recursive, you are able to reproduce the original correlation matrix from the paths. Below is a correlation matrix among Z1 (Gender), Z2 (Race), Z3 (S’s education), Z4 (S’s occupation), and Z5 (S’s present income). Missing in this matrix are all correlations with Z5. Using the values of paths in the above path model, calculate these four missing correlations (identified in the matrix with question marks). Show your work. Z1 Z2 Z3 Z4 Z1 1.0 Z2 .30 1.0 Z3 .25 .15 1.0 Z4 .30 .35 .40 1.0 Z5 ? ? ? ? Z5 1.0 c. The model does not include coefficients for the residual paths to S's education, S's occupation, or S's income. Calculate these. 1 d. Calculate the total, direct, and indirect effects of the exogenous variables and endogenous variables on the endogenous variables. Present these effects in a table that shows that the total effects equal the sums of the direct and indirect effects. e. Delete the paths from race to S's education and from gender to S's present income and draw the revised path diagram. (In Lab 12 this restricted model will be compared to the fully recursive model for goodness of fit.) Some help for part e: The following program illustrates how to regress Y on W and X when you only know that rYW = .3 , rYX = .25 , and rWX = .2 in a sample of n = 100 . Please note that you will need to modify this program to obtain some of the paths called for in part e: matrix data vars=y,w,x / contents=mean sd corr / n=100. begin data. 0 0 0 1 1 1 1.00 .30 1.00 .25 .20 1.00 end data. regression matrix=in(*) / var=y,w,x / dep=y / enter. 2. There are two conflicting theories on the development of Post-Traumatic Stress Syndrome (a psychological disorder in which American soldiers who fought in the Iraq War have later difficulties in adjusting to civilian life). According to one theory (Theory A), a soldier's strict family background (X1) prevents the soldier from having a relativistic world view (X2). The theory continues to argue that without a relativistic world view, soldiers who fought in Iraq develop symptoms of the syndrome (X3). A less parsimonious theory (Theory B) is entirely in agreement with Theory A, but also notes that people who grow up in strict families often have symptoms of the syndrome. (For example, people from strict family backgrounds are known to have sudden uncontrollable outbursts of anger). You collect data on 120 veterans of the War in Iraq and (using factor analysis) you develop the following three measures: X1 = Strictness of family background (high scores are strict). X2 = Relativity of world view (high scores are relativistic). X3 = Degree to which Post-Traumatic Stress Syndrome is manifest (high scores reflect the greatest manifestation of the syndrome) 2 Correlations among these variables are . . . X1 X2 X3 X1 1.0 -.30 .10 X2 -.30 1.0 .40 X3 .10 .40 1.0 a. Draw a path model for Theory A that includes all path and residual coefficients. (Show your calculations!) b. Draw a path model for Theory B that includes all path and residual coefficients. (Show your calculations!) c. Construct a table of effects for EACH path model. 3. An industrial psychologist is investigating the determinants of job performance. She has data on four variables from a sample of 100 married employees of the American Widget Company. The variable names are . . . PERFORM: COMMIT: HAPPY: KIDS: High scores indicate high job performance High scores indicate high job commitment High scores indicate high marital happiness Number of children under 18 years of age in the employee's family Correlations among these variables are . . . PERFORM COMMIT HAPPY KIDS PERFORM 1.00 .65 .20 .30 COMMIT .65 1.00 -.25 -.20 HAPPY .20 -.25 1.00 -.03 KIDS .30 -.20 -.03 1.00 a. It is well-established in previous research that job commitment enhances (i.e., increases) job performance. It is also well-established that job commitment is less common among employees with happy marriages than with unhappy marriages. (This is due to happily married employees' displacement of commitment from their jobs to their marriages.) Using data presented in the correlation matrix, draw a path model that depicts these well-established relations among PERFORM, COMMIT, and HAPPY. (Be sure to include the numerical values for all relevant paths and for all residual coefficients as well as to indicate how you arrived at these values.) 3 b. What has not been established in previous research is whether the number of an employee's children has an effect on the well-established process described in part a. In particular, the industrial psychologist theorizes that, although job performance is likely to be greater among employees with increasingly larger numbers of children (due to the increasing feeling of family responsibility that parenthood brings), the total effect of employees' number of children will be of a decrease in job performance, because children (like spouses in happy marriages) entice displacement of commitment from employees' jobs to their families. To test the role of KIDS in the causal relations among HAPPY, COMMIT, and PERFORM, the path model in part a was extended to a model that included statistics from output generated by the following two SPSS commands: regression vars=COMMIT,HAPPY,KIDS / dep=COMMIT / enter regression vars=PERFORM,COMMIT,KIDS / dep=PERFORM / enter Part of the output from the first regression is . . . ------------------ VARIABLES IN THE EQUATION -----------------VARIABLE B SE B BETA KIDS HAPPY (constant) -4.153738 -.512461 89.055150 1.921351 .192135 12.408583 -.207687 -.256231 Part of the output from the second regression is . . . ------------------ VARIABLES IN THE EQUATION -----------------VARIABLE B SE B BETA KIDS COMMIT (constant) 13.437500 1.109375 -42.343750 1.928730 .096436 7.007529 .447917 .739583 In her extended path model, the researcher decided not to specify a causal direction in the relation between KIDS and HAPPY. Draw the path model that depicts this unanalyzed relation, the causal relations that the researcher estimated in the above regressions, and the well-established causal relations depicted in the path model drawn in part a. (Again, be sure to include the numerical values for all relevant paths and for all residual coefficients as well as to indicate how you arrived at these values.) 4 c. Based on the path model drawn in part b, what is the direct effect of KIDS on PERFORM? (Be sure to indicate how you arrived at your answer.) d. Based on the path model drawn in part b, what is the indirect effect of KIDS on PERFORM? (Be sure to indicate how you arrived at your answer.) Below please find R and SAS code for problem 1.e.: # R # Code: nvars <- 3 #----Enter the number of variables----# corrs <- matrix(c (1.00, 0.30, 0.25, #----Enter the correlation matrix with----# 0.30, 1.00, 0.20, #----Y in column 1 and row 1.----# 0.25, 0.20, 1.00) , ncol=nvars, byrow=T) R <- corrs[2:nvars,2:nvars] #----Create matrix R of corrs excluding column 1 and row 1----# r <- corrs[1,2:nvars] #----Create vector r of corrs in first row excluding the first value----# beta.hats <- solve(R) %*% r #----The inverse of R multiplied by r----# beta.hats * SAS * Code; PROC IML; corrs = {1.00 0.30 0.25, /*****Enter correlation matrix with Y in 0.30 1.00 0.20, column 1 and row 1*****/ 0.25 0.20 1.00}; Rbig = corrs[{2 3}, {2 3}]; /*****Create submatrix R consisting of rows 2-3 and columns 2-3*****/ rsmall = corrs[{1}, {2 3}]; /*****Create submatrix r consisting of row 1, columns 2-3*****/ Rbig_inv = inv(Rbig); /***** Inverse of Rbig *****/ rsmall_trans = (rsmall)`; /***** Inverse of rsmall *****/ betas = rbig_inv*rsmall_trans; /***** Matrix muliplication *****/ 5 PRINT betas; RUN; 6