This past week, I have been using the data I have collected in Stata to see what I could find. I made a model for estimating how many uiuc transfer students an IL community college would have, but every model had Parkland Community College as an outlier, so I made one for all schools besides Parkland. After that, I tried to create a model of IL community colleges' graduation and transfer rates based on school characteristics. My transfer model without Parkland had an Adjusted R^2 of around .85, while the school performance model had an adjusted R^2 around .60, so I was relatively happy with them.
Then I tried to use the data to check how the Engineering Pathways program effects engineering transfers to UIUC, with much less success. First, I tried to model which schools would qualify for the program, but my best model predicted less than half of the schools. This suggests that school qualification for the program is based factors other than school characteristics, perhaps state politics or personal relationships. And while Pathways schools tend to have more engineering transfers, the effect is small, averaging a little over 2 more transfers, and Pathways status has no effect on percent of engineering transfers to uiuc out of total uiuc transfers. If I were to do a section on Pathways in my final paper, it would probably reflect more on how ineffective it is.
My goal now is to start writing up my analysis in paper form, and hopefully get a rough draft within the next week or two. I'll let you know its progress in future posts.
Apropos my comments to your previous post, you should not just have such a model but also try to understand your data better, apart from the Parkland is different answer. Some much simpler analysis on correlations across variable would help here and either validate your model or show how to modify it suitably.
ReplyDeleteI know we're getting down to crunch time, so at some point you are just going to have to wrap it up, but I will continue to comment as if that is not an issue. Your understanding of the results you are trying to get is my concern.