Homework 16

Objectives:  Analyze Qualitative Choice Models using the AACSB.XLS file.

Data Background:

The data file AACSB.XLS contains data for 404 graduate business school programs on six variables, the names of which are contained in the first record of the file. The variable AACSB is a 1,0 dummy variable that indicates whether or not the American Assembly of Collegiate Schools of Business (AACSB) accredits the business school program.    Three other variables indicate the number of students enrolled (SIZE), the average GMAT score of enrollees, and the proportion of faculty with doctorates (FACPHD).  Two additional 1,0 dummy variables indicate whether the program granted a doctoral business degree (PHD), and whether the program was a public university (STATE).  The data for this file come from a study of accrediting practices published by Jantzen and Pendleton ("Preferences of the American Assembly of Collegiate Schools of Business," Journal of Education for Business, Vol. 70, Sept/Oct 1994,  pp. 6-11).  About two-fifths of all graduate business programs were accredited at the time the data were collected (1992).

Assignment:

a.   Generate descriptive statistics for the AACSB, SIZE, GMAT, FACPHD, PHD and STATE variables using EALimdep (click here for help).   What do they tell us about the average values and dispersion of the five variables.  Discuss in the specific terms in which each variable is measured.

b.  What kinds of problems are likely to arise if the AACSB variable is used as a dependent variable in an OLS regression?

c.  Use EALimdep to estimate an OLS (Linear Probability) regression model explaining which schools were accredited by the AACSB.  Use the SIZE, GMAT, FACPHD, PHD and STATE variables as explainers.  Interpret the estimated coefficients.  Click on the output tab and the <keep predictions as variable> button and give the predictions a name like PFIT.

d.  Conduct appropriate t-tests on the OLS regression coefficients.  Which variables significantly effect the probability of being AACSB accredited?

e.  Evaluate the goodness of fit of the OLS regression model.

f.  Test whether the OLS results are influenced by heteroskedasticity.

g.  Are the OLS predicted probabilities less than zero or greater than 1?  Click on <DataDescription> and generate a histogram for the predicted probabilities that you've already saved as PFIT.

h.  Use EALimdep to estimate a logit regression model explaining which schools were accredited by the AACSB (click here for help).  Use the SIZE, GMAT, FACPHD, PHD and STATE variables as explainers.  Is the logit estimator superior to the OLS estimator?  How so?

i.  Conduct appropriate t-tests on the logit regression coefficients.  Which variables significantly effect the probability of being AACSB accredited?

j.  Transform the logit regression coefficients into coefficients that measure the effect of each explainer on the probability of being accredited.  Interpret the transformed coefficients.

k.  Evaluate the goodness of fit of the logit regression model by comparing the predicted accreditation status of each school with its actual status.  Also examine the pseudo- R squared.

l.  Conduct a likelihood ratio test on the overall LOGIT model.

m.   Conduct a likelihood ratio test on whether or not the coefficients on the FACPHD and the PHD variables are both zeros.