my last report amended (attached)

Date: Apr 15, 2004

my last report amended (attached)

Date: Apr 15, 2004

lsu social science grad student--see my responses to your analyses on your original thread.

Date: Apr 15, 2004

Iddad:

The AdmissIndex is a predicted value for the U.S. News admissions standards index for a university. You noted that the 2001 to 2002 drop was small, but the 2002 to 2003 movement was much larger.

I certainly didn't mean to bring a causal message, although looking back at some of the phrases, I could've used softer language in this regard. There are no significance tests of the differences, although they could be done (I didn't spend that much time with it).

I've talked to some people at USM, and their feeling all along is that Thames has been lowering admissions standards (I guess from classroom experiences).

Date: Apr 15, 2004

come on lsu social science graduate student--i'm teaching undergraduates right now about significance testing. those numbers look nice but what about tests of significance? this is technical for most of this board but let's not try to draw conclusions unwarranted by the data. what should we ethically conclude?

Date: Apr 15, 2004

If what the faculty members have told me is correct, then the results are not inconsistent with their experiences. That's all I meant in my last post. The numbers are right in front of you (on the post) --- do a significance test if you want.

Date: Apr 15, 2004

sorry--there is not enough data to test for significant differences in what you have provided. come on! as sophisticated as your analyses are, you know that. the numbers provided don't allow me to make such a comparison. to allude to conversations you have with people at USM you might not be objective in your analyses. you could provide me with your data set and i could replicate.

Date: Apr 15, 2004

Here's what you can say. The standard deviation (amongthose indexes) is 0.054, and the difference in the index when Thames took over and its most recent value is 0.123, which is close to a decrease of 2.3 standard deviations.

Date: Apr 15, 2004

Here's what you can say. The standard deviation (amongthose indexes) is 0.054, and the difference in the index when Thames took over and its most recent value is 0.123, which is close to a decrease of 2.3 standard deviations.

I don't want what I can say. your statement is so vague ("among those indexes") that it has no validity. i want legitimate analyses. your non-response about your data set is beginning to tell me something about your analyses.

Date: Apr 15, 2004

Look man, here's what I did. I took the correlation between a university's admissions standard from US News and its freshman retention rate, using about 100 observations. From that regression, I generated 5 admission index predictions for USM since the Clarion Ledger had only five years of freshman retention rates. You don't need my data, you have my indexes. You say you can't do a significance test with those indexes, yet I did one for you. You could run a regression with those indexes if you wanted to. There are five of them, have you ever heard of the rule of 5?

I told you what I did even though the report post did as well.

Date: Apr 15, 2004

I don't want what I can say. your statement is so vague ("among those indexes") that it has no validity. i want legitimate analyses. your non-response about your data set is beginning to tell me something about your analyses.

What non-response? Didn't you read the report? What's so vague about "among those indexes"? They're right in front of you.

Date: Apr 15, 2004

Iddad, are you coming back?

Date: Apr 15, 2004

your non-response about your data set is beginning to tell me something about your analyses.

You don't have to hurl these sentences my way. I'm answering your questions.

Date: Apr 15, 2004

Can I log off?

Date: Apr 15, 2004

LSU Grad Student, did you happen to see what I reminded everyone someone said about iddad?

Fire Shelby · Status: Offline

quote:
Originally posted by: LSU Grad Student
"Can I log off?"

I want to reiterate: Thank you VERY much for posting this study. I hope you won't feel discouraged from posting again!

Date: Apr 16, 2004

i'm back. still have my two original concerns.

1. inferring causality from correlational data. you address a bit, but ultimately admit your conclusions about ST's role are based on conversations with people at USM not your data. no problem, but the data aren't driving your conclusion.

2. there is a potential range restriction problem in the data set. this you haven't addressed, and it's a bit of a technical issue, but it makes a potentially major difference in your statistical analyses and the conclusions you want to draw. as i noted somewhere in one of the two threads, Ole Miss may be getting the higher ACT students. Their freshman class is growing in major ways (from the clarion ledger data), even jumping well past MSU. if they are taking away the higher ACT students because of their Barksdale honor's program and the new Phi Beta Kappa chapter, then other universities' data on retention may look bad because the other universities are dealing with lower (on average) ACT students, not because of admission's policies (your conclusion) but because of a lower ACT student body (alternative conclusion).

Date: Apr 16, 2004

Iddad---

I don't doubt the Ole Miss hypothesis you indicate. Then, it would be that both factors (the Ole Miss thing and the admissions policy thing) are playing a role.

Date: Apr 16, 2004

Once again I do not state causality. If you get to tell me what I'm thinking and saying then we can't have a real discussion.

I said there is a correlation between the admissions policy and the fresh retention rates.

Date: Apr 16, 2004

It's seems funny that you get to glance at the data and conclude that Ole Miss' actions are CAUSING USM's freshman numbers to vary but when I got out do some statistical testing, and then ask some USM people is their casual empiricism supports it, I'm committing all sorts of statistical fatal flaws. That's interesting .....

Fire Shelby · Status: Offline

quote:
Originally posted by: lddad
"i'm back. still have my two original concerns. 1. inferring causality from correlational data. you address a bit, but ultimately admit your conclusions about ST's role are based on conversations with people at USM not your data. no problem, but the data aren't driving your conclusion. 2. there is a potential range restriction problem in the data set. this you haven't addressed, and it's a bit of a technical issue, but it makes a potentially major difference in your statistical analyses and the conclusions you want to draw. as i noted somewhere in one of the two threads, Ole Miss may be getting the higher ACT students. Their freshman class is growing in major ways (from the clarion ledger data), even jumping well past MSU. if they are taking away the higher ACT students because of their Barksdale honor's program and the new Phi Beta Kappa chapter, then other universities' data on retention may look bad because the other universities are dealing with lower (on average) ACT students, not because of admission's policies (your conclusion) but because of a lower ACT student body (alternative conclusion). "

1. You and I have been through this very argument before. LSU Grad's inference is just that--an inference. It doesn't change the fact that he/she has isolated a positive correlation that is very significant.

2. Perhaps the higher ACT students wouldn't HAVE to consider Ole Miss if Shelby had not dismantled the Honors College, which was nationally renown, at USM, and of which I am a proud graduate.