|
|
|
||||||||||
 |
||||||||||||
 |
||||||||||||
PA 784
Fall 2002
First: Ensure that SPSS is configured to give you your program commands on your output: Go to Edit/Preferences/Options/Draft Viewer, and make sure that “Display Commands in Log” is checked.
The purpose of the SSS-TRIO program is to increase the likelihood that students at risk for dropping out of college are retained and graduate. In this particular analysis, you will test the following piece of the overall program theory:
Recruitment into TRIO Participate in program activites Increase in GPA
(1) Examine whether each of the activities (tutoring, advising, workshops) are (independently) having their intended effect of increasing GPA. We will use the Fall semester GPA as the “pretest” and the Spring semester GPA as the “posttest”. Use data file triostds.sav (available in PA 784 folder).
a) First, find out the average number of advising sessions and hours of tutoring students have taken (in the Spring semester) and the number who attended the Spring workshop.
b) Then, use regression analysis to assess the extent to which (if at all) these activities improved students’ GPAs in the Spring semester. What do you conclude?
(2) Assess program impact by using a comparison group as the measure of the counterfactual. For this exercise, use the file trio&ctl.sav (in PA 784 folder)
a) First, Look at how the treatment (TRIO) and comparison (“regular EOP”) students score on the pre and post-tests (again, using Fall GPA as the pre-test and Spring GPA as the post-test. The “program” variable identifies the group each student was in.
b) Then use regression analysis to determine if the program appears to have been effective in raising students’ GPAs. What do you conclude?
c) Run the regression again, including any variables that you think should be controlled for. How do control variable(s) affect your results?
The purpose of “managing diversity” programs is to create a more equitable work environment for women and people of color. Here you will assess whether it has done so at two federal agencies: the National Oceanic and Atmospheric Administration (NOAA) and National Institutes of Health (NIH). Ideally, these should include many more data points than we have (particularly post-intervention), but we will make do.
1) Open the file NOAAquit.sav (in PA 784 folder). The dependent variable in this analysis is a ratio representing the extent to which people of color were promoted in proportion to their representation in the population and in proportion to promotions among white employees. Basically, a ratio of 1.0 means they were promoted at the same rate as everyone else; a ratio below 1.0 means they were promoted disproportionately less. So, we assume that if the managing diversity program is having an impact, it should increase the promotion ratio for people of color. NOAA implemented its program in 1994.
Run the necessary regression to estimate program impact. What was happening to the quit ratio before the program was implemented? What was the immediate impact of the program? What happened to the quit ratio afterwards?
2) This time, you will set up your own data file, using the data in the attached table below. These are data from NIH representing the extent to which African Americans are disproportionately dismissed from their jobs (calculated the same was as quits above). Therefore, this time we assume an effective diversity program would decrease the dismissal ratio for African Americans. NIH’s diversity program was implemented in 1995.
a) Add the three variables: the counter for time, the treatment, and the interaction between the counter and treatment.
b) Run a regression to estimate program impact. What was happening to the dismissal ratio before the program was implemented? What was the immediate impact of the program? What happened to the dismissal ratio afterwards?
Dismissal
Ratios at NIH
|
Year |
Dismissal
Ratio |
|
1988 |
2.138 |
|
1989 |
3.374 |
|
1990 |
2.288 |
|
1991 |
2.681 |
|
1992 |
1.706 |
|
1993 |
2.987 |
|
1994 |
3.574 |
|
1995 |
2.852 |
|
1996 |
2.057 |
|
1997 |
2.64 |
|
1998 |
2.4 |
|
1999 |
1.273 |
|
1988 |
2.138 |
In this exercise, students with GPAs above 1.5 were assigned to the program and those with GPAs below served as the comparison group. The outcome measure is their GPA after completing the mentoring program. Use the datafile mentrd.sav (PA 784 folder).
1) Run a regression to estimate program impact. What do you find?
2) Remember that this design requires the slopes for the two groups to be the same (that is, the relationship between the assign variable and outcome measure to be the same). Test whether this is the case by running separate regressions for each group.
What do you find? Can we be confident of our impact estimate?