Interesting vid on urm lsats and admissions

(Rankings, Profiles, Tuition, Student Life, . . . )
User avatar
alisyn313
Posts: 50
Joined: Tue Jun 23, 2009 5:13 pm

Re: Interesting vid on urm lsats and admissions

Postby alisyn313 » Thu Feb 25, 2010 1:50 am

Wow....very interesting.

User avatar
ravens20
Posts: 192
Joined: Mon Oct 19, 2009 8:17 pm

Re: Interesting vid on urm lsats and admissions

Postby ravens20 » Fri Feb 26, 2010 5:27 am

ogman05 wrote:This is a long clip but definitely proves and provides some good info. Only 125 above 170.

http://www.youtube.com/watch?v=_7_xHsce ... tube_gdata

first thing I will argue is that he mentions that any item that favored blacks was removed basically saying the items that did not favor them are left in (I kno LR fail but work with me). This leads to the argument that the test is created for a certain group and is easier for hem to score well. The test is still predictive for minority success relative to other minorities however. ThatS why there are different score bands for different groups in admissions.


First of all, it was a very interesting and informative video, which I thoroughly enjoyed. Thank you for posting it.

But after watching this two times to try and find out where Professor Johnson makes the argument that you mentioned in your post, I think you are misquoting him and ultimately missing his point. In fact, he is asserting the opposite of what you attributed to him regarding the LSAT being biased.

He says that LSAC tests every potential question for bias and throws out those that favor people of ANY race, gender, etc. You are referring to the example he gives where a question that favored African Americans was thrown out...but that is just one example of many similar questions that were thrown out for being biased in favor of one group or another (he points out that there was another one that favored whites). In fact, the example you pointed out was when he was making the argument that there was little bias in the test; throwing out questions that favor anyone demonstrates that the test is fair. He also never makes the claim that the test was created for a certain group. His entire point was that there wasn't any bias in the test and that there is no evidence to suggest that the test favors non-URMs. He repeatedly says that he has no idea why there is a gap between whites and African Americans at every socio-economic level, but that the LSAC folk ensure through rigorous testing that such a gap couldn't be due to the test being biased (he even points out the same standard deviation difference occurs in nearly every standardized test). In fact he goes out of his way to show that the LSAT is a good test and says so explicitly on several occasions.

The point of the lecture is not to explain WHY there is a gap between LSAT scores, which he says that he cannot do (and specifically points out that it is not socio-economic status or bias in the LSAT), but why African-Americans are not being accepted at law schools in larger numbers when the total number of seats at law schools has increased since 1993. To answer this he gives several theories: 1) deans changing acceptance policies based on incorrect assumptions of the outcome of the Grutter v. Bollinger case; 2) deans not wanting to make acceptance decisions based on race; 3) increased emphasis on LSAT scores due to US News rankings; 4) foolish deans of admission not realizing that schools can accept students of any LSAT below their desired 25th percentile scores; 5) admissions officers not wanting to put low scoring kids at a disadvantage, as the LSAT offers statistically significant (although still very minor) predictive value of performance in law school; 6) that African-Americans are applying to the wrong schools or are getting bad application advice. These are his arguments, not that the LSAT is biased.
Last edited by ravens20 on Fri Feb 26, 2010 6:26 am, edited 6 times in total.

User avatar
flyingpanda
Posts: 824
Joined: Mon Oct 19, 2009 5:32 am

Re: Interesting vid on urm lsats and admissions

Postby flyingpanda » Fri Feb 26, 2010 5:37 am

ravens20 wrote:
ogman05 wrote:
Exactly my point. Without this it would got back to how it was pre-1940's with elite UG's. They would need some other measure of performance. After hearing him talk I am convinced that there is a bias inherent within the test. Although it is predictive, URM's do worse on average. He even mentioned that there was an item that urms distinctively did better on. If the entire test were made up of these "items" wouldnt it jsut change the playing field and make it more level. Then non-urms would be getting boosted. haha. Now understanding why the lsat has such little predictive quality for urms when compared to other nonurms makes sense. a 150 become a 155 or 160. A 160 a 165 or 170. Makes sense.


I'm not sure how you came to the bolded conclusion from what was said in this lecture. Professor Johnson was basically asserting the opposite off what you just stated.

Professor Johnson went out of the way to point out that there was NO bias in the test. He mentions several times that the LSAT is fair in his eyes. He talked about how each question goes through a test from people of every race/ethnicity, gender, etc to determine whether there is any bias against any of them. He speaks about questions that were thrown out for favoring non-URMs and also one in particular that favored African Americans...neither should be allowed in a fair test and neither is. His entire point was that there wasn't any bias in the test and that there is no evidence to suggest that the test favors non-URMs. He repeatedly says that he has no idea why there is a gap between whites and African Americans at every socio-economic level, but that the LSAC folk ensure that such a gap couldn't be due to the test being biased (he even points out the same standard deviation difference occurs in nearly every standardized test).


Correct me if I'm wrong, but I thought the URM under-performance (don't know what else to call it) only comes out in the overall test. It seems like the experimental section tries to find individual questions that don't have an inherent bias.

Also Professor Johnson pointed out that the LSAT seems more predictive for URMs than for non-URMs.

In any case, the video was pretty fascinating and I'm surprised that I actually watched the whole thing.

User avatar
ravens20
Posts: 192
Joined: Mon Oct 19, 2009 8:17 pm

Re: Interesting vid on urm lsats and admissions

Postby ravens20 » Fri Feb 26, 2010 6:10 am

aznflyingpanda wrote:
Correct me if I'm wrong, but I thought the URM under-performance (don't know what else to call it) only comes out in the overall test. It seems like the experimental section tries to find individual questions that don't have an inherent bias.

Also Professor Johnson pointed out that the LSAT seems more predictive for URMs than for non-URMs.

In any case, the video was pretty fascinating and I'm surprised that I actually watched the whole thing.


Yea he did mention that the predictive power for URMs was less. But he also pointed out how LSAC tried to make the test as unbiased as possible, so the test was not really biased (after all the same gap apparently occurs in other standardized tests as well). What I was getting at is that Professor Johnson isn't trying to explain the gap in the LSAT or other standardized tests (he even jokes that if he could, he would get a Nobel Prize), but why more seats have been made available for prospective law students but the number of admitted African-Americans has remained the same.

And I totally agree...the lecture was surprisingly engaging. I actually watched it twice.

User avatar
dailygrind
Posts: 19639
Joined: Wed Oct 22, 2008 11:08 am

Re: Interesting vid on urm lsats and admissions

Postby dailygrind » Fri Feb 26, 2010 11:11 am

If you start watching at 12:35, he says specifically that the predictive validity for African Americans is higher than it is for Caucasians. He says this to counter the idea that the test is biased against African Americans.

He doesn't go into what the coefficients are, which means that it's possible that it's more accurate for African Americans and a 160 for an African American is equivalent to a 165 for a Caucasian, but I have a hard time believing that he wouldn't mention that if it were the case.

User avatar
ogman05
Posts: 266
Joined: Sun Apr 19, 2009 5:02 pm

Re: Interesting vid on urm lsats and admissions

Postby ogman05 » Fri Feb 26, 2010 11:27 am

ravens20 wrote:
ogman05 wrote:This is a long clip but definitely proves and provides some good info. Only 125 above 170.

http://www.youtube.com/watch?v=_7_xHsce ... tube_gdata

first thing I will argue is that he mentions that any item that favored blacks was removed basically saying the items that did not favor them are left in (I kno LR fail but work with me). This leads to the argument that the test is created for a certain group and is easier for hem to score well. The test is still predictive for minority success relative to other minorities however. ThatS why there are different score bands for different groups in admissions.


First of all, it was a very interesting and informative video, which I thoroughly enjoyed. Thank you for posting it.

But after watching this two times to try and find out where Professor Johnson makes the argument that you mentioned in your post, I think you are misquoting him and ultimately missing his point. In fact, he is asserting the opposite of what you attributed to him regarding the LSAT being biased.

He says that LSAC tests every potential question for bias and throws out those that favor people of ANY race, gender, etc. You are referring to the example he gives where a question that favored African Americans was thrown out...but that is just one example of many similar questions that were thrown out for being biased in favor of one group or another (he points out that there was another one that favored whites). In fact, the example you pointed out was when he was making the argument that there was little bias in the test; throwing out questions that favor anyone demonstrates that the test is fair. He also never makes the claim that the test was created for a certain group. His entire point was that there wasn't any bias in the test and that there is no evidence to suggest that the test favors non-URMs. He repeatedly says that he has no idea why there is a gap between whites and African Americans at every socio-economic level, but that the LSAC folk ensure through rigorous testing that such a gap couldn't be due to the test being biased (he even points out the same standard deviation difference occurs in nearly every standardized test). In fact he goes out of his way to show that the LSAT is a good test and says so explicitly on several occasions.

The point of the lecture is not to explain WHY there is a gap between LSAT scores, which he says that he cannot do (and specifically points out that it is not socio-economic status or bias in the LSAT), but why African-Americans are not being accepted at law schools in larger numbers when the total number of seats at law schools has increased since 1993. To answer this he gives several theories: 1) deans changing acceptance policies based on incorrect assumptions of the outcome of the Grutter v. Bollinger case; 2) deans not wanting to make acceptance decisions based on race; 3) increased emphasis on LSAT scores due to US News rankings; 4) foolish deans of admission not realizing that schools can accept students of any LSAT below their desired 25th percentile scores; 5) admissions officers not wanting to put low scoring kids at a disadvantage, as the LSAT offers statistically significant (although still very minor) predictive value of performance in law school; 6) that African-Americans are applying to the wrong schools or are getting bad application advice. These are his arguments, not that the LSAT is biased.


I understand all of what you are trying to say and was not trying to misquote him. I am only saying that according to "previous tests" there are no bias in the LSAT. But what are those tests based on. How do you control for bias? It is impossible unless you are using historical data which could in and of itself be biased. According to his statisticians there is no bias but how can they prove this I ask you? HISTORICAL data. That is how you run the regressions and other statistical calculators that they used. I took stats and upper level courses on this. If the data is biased your answer will be. End of story.

If 100 kids took a test and 20 were a certain group and all did worse. How do they test and control for that. They would either claim bias or compare it to a future test where 20 kids of that same group also did worse. There might not be any bias but there may. Within a certain subgroup the test can be predictive. But just like comparing across law schools, comparing across subgroups are more difficult. This is the part that he argues is unexplainable imo.

User avatar
ogman05
Posts: 266
Joined: Sun Apr 19, 2009 5:02 pm

Re: Interesting vid on urm lsats and admissions

Postby ogman05 » Fri Feb 26, 2010 11:34 am

dailygrind wrote:If you start watching at 12:35, he says specifically that the predictive validity for African Americans is higher than it is for Caucasians. He says this to counter the idea that the test is biased against African Americans.

He doesn't go into what the coefficients are, which means that it's possible that it's more accurate for African Americans and a 160 for an African American is equivalent to a 165 for a Caucasian, but I have a hard time believing that he wouldn't mention that if it were the case.


Yes I agree. WIthin the group it is more predictive. Jsut not across groups. This does not mean that it is not biased against African Americans compared to other groups. Just that it has predictive nature. That I did not agree on. The fact that it can be predictive =/= no bias.

ie. If I have 100 people take a test on country music. And 20 are black and 80 were white. The black test takers will probably due worse on average than say their white counterparts. This would be because they are not listening to that music on average so have more difficulty answering questions about it. However inside that subgroup(blacks) there may be a predictive quality about their success in music school. Maybe even greater so than the white test takers given that their scores were lower. This predictive quality exist while there would be a clear bias in the test. Therefore, predictive quality =/= no bias.

User avatar
ogman05
Posts: 266
Joined: Sun Apr 19, 2009 5:02 pm

Re: Interesting vid on urm lsats and admissions

Postby ogman05 » Fri Feb 26, 2010 11:35 am

ravens20 wrote:
aznflyingpanda wrote:
Correct me if I'm wrong, but I thought the URM under-performance (don't know what else to call it) only comes out in the overall test. It seems like the experimental section tries to find individual questions that don't have an inherent bias.

Also Professor Johnson pointed out that the LSAT seems more predictive for URMs than for non-URMs.

In any case, the video was pretty fascinating and I'm surprised that I actually watched the whole thing.


Yea he did mention that the predictive power for URMs was less. But he also pointed out how LSAC tried to make the test as unbiased as possible, so the test was not really biased (after all the same gap apparently occurs in other standardized tests as well). What I was getting at is that Professor Johnson isn't trying to explain the gap in the LSAT or other standardized tests (he even jokes that if he could, he would get a Nobel Prize), but why more seats have been made available for prospective law students but the number of admitted African-Americans has remained the same.

And I totally agree...the lecture was surprisingly engaging. I actually watched it twice.


Look at the bold. trying=/= perfect success. Just means that they tried. They could be doing their best and failing miserably. LR fail

User avatar
dailygrind
Posts: 19639
Joined: Wed Oct 22, 2008 11:08 am

Re: Interesting vid on urm lsats and admissions

Postby dailygrind » Fri Feb 26, 2010 11:39 am

ogman05 wrote:I understand all of what you are trying to say and was not trying to misquote him. I am only saying that according to "previous tests" there are no bias in the LSAT. But what are those tests based on. How do you control for bias? It is impossible unless you are using historical data which could in and of itself be biased. According to his statisticians there is no bias but how can they prove this I ask you? HISTORICAL data. That is how you run the regressions and other statistical calculators that they used. I took stats and upper level courses on this. If the data is biased your answer will be. End of story.

If 100 kids took a test and 20 were a certain group and all did worse. How do they test and control for that. They would either claim bias or compare it to a future test where 20 kids of that same group also did worse. There might not be any bias but there may. Within a certain subgroup the test can be predictive. But just like comparing across law schools, comparing across subgroups are more difficult. This is the part that he argues is unexplainable imo.


So your contention is that the historical data he's been using is biased.

In order for the bias to push down African American scores now, the historical data would have to measure African Americans that are, as a group, more intelligent than the African Americans who are currently being measured. So they show no difference on the tests, everything is good, then along comes the present generation who doesn't measure up, causing bias in the testing. My mind must be inflexible without caffeine, as this would be a strange thing to argue, and you must mean something else. How would you explain the situation in which the historical data was biased?

User avatar
dailygrind
Posts: 19639
Joined: Wed Oct 22, 2008 11:08 am

Re: Interesting vid on urm lsats and admissions

Postby dailygrind » Fri Feb 26, 2010 11:41 am

ogman05 wrote:
dailygrind wrote:If you start watching at 12:35, he says specifically that the predictive validity for African Americans is higher than it is for Caucasians. He says this to counter the idea that the test is biased against African Americans.

He doesn't go into what the coefficients are, which means that it's possible that it's more accurate for African Americans and a 160 for an African American is equivalent to a 165 for a Caucasian, but I have a hard time believing that he wouldn't mention that if it were the case.


Yes I agree. WIthin the group it is more predictive. Jsut not across groups. This does not mean that it is not biased against African Americans compared to other groups. Just that it has predictive nature. That I did not agree on. The fact that it can be predictive =/= no bias.

ie. If I have 100 people take a test on country music. And 20 are black and 80 were white. The black test takers will probably due worse on average than say their white counterparts. This would be because they are not listening to that music on average so have more difficulty answering questions about it. However inside that subgroup(blacks) there may be a predictive quality about their success in music school. Maybe even greater so than the white test takers given that their scores were lower. This predictive quality exist while there would be a clear bias in the test. Therefore, predictive quality =/= no bias.


What you were just talking about is what I was getting at with the coefficients talk. Like I said then, it's conceivable that such is the case - he did not explicitly talk about it. However, I find it difficult to believe that in a 42 minute lecture in which he covered in detail so much of the admissions process and how it may be biased, he would not have mentioned that if it were the case.

User avatar
dailygrind
Posts: 19639
Joined: Wed Oct 22, 2008 11:08 am

Re: Interesting vid on urm lsats and admissions

Postby dailygrind » Fri Feb 26, 2010 11:46 am

For the record, if there were a bias against African Americans while predictive validity was higher for them, it would just mean that a 160 for an African American was equivalent to a 164 or for a Caucasian, or some other number. It would be extremely easy for me to pick up in statistical analysis, and I definitely do not have a PhD specializing in this field.

09042014
Posts: 18282
Joined: Wed Oct 14, 2009 10:47 pm

Re: Interesting vid on urm lsats and admissions

Postby 09042014 » Fri Feb 26, 2010 11:48 am

ogman05 wrote:
I understand all of what you are trying to say and was not trying to misquote him. I am only saying that according to "previous tests" there are no bias in the LSAT. But what are those tests based on. How do you control for bias? It is impossible unless you are using historical data which could in and of itself be biased. According to his statisticians there is no bias but how can they prove this I ask you? HISTORICAL data. That is how you run the regressions and other statistical calculators that they used. I took stats and upper level courses on this. If the data is biased your answer will be. End of story.

If 100 kids took a test and 20 were a certain group and all did worse. How do they test and control for that. They would either claim bias or compare it to a future test where 20 kids of that same group also did worse. There might not be any bias but there may. Within a certain subgroup the test can be predictive. But just like comparing across law schools, comparing across subgroups are more difficult. This is the part that he argues is unexplainable imo.


What he said was that the scoring differential exists on every single question. If it was a bias issues you'd likely see some questions be more biased that others, but they the scoring difference is the same on almost all questions. So he concluded that the LSAT is not a biased standardized test.

The difference in testing results continues into law school. They do as well as their LSAT suggests they will. That is another argument that the LSAT is an unbiased test.

Whatever is causing the difference is not likely a bias on the test, or on law school exams. Like he said, he doesn't know why it happens.

User avatar
ogman05
Posts: 266
Joined: Sun Apr 19, 2009 5:02 pm

Re: Interesting vid on urm lsats and admissions

Postby ogman05 » Fri Feb 26, 2010 11:54 am

dailygrind wrote:
ogman05 wrote:I understand all of what you are trying to say and was not trying to misquote him. I am only saying that according to "previous tests" there are no bias in the LSAT. But what are those tests based on. How do you control for bias? It is impossible unless you are using historical data which could in and of itself be biased. According to his statisticians there is no bias but how can they prove this I ask you? HISTORICAL data. That is how you run the regressions and other statistical calculators that they used. I took stats and upper level courses on this. If the data is biased your answer will be. End of story.

If 100 kids took a test and 20 were a certain group and all did worse. How do they test and control for that. They would either claim bias or compare it to a future test where 20 kids of that same group also did worse. There might not be any bias but there may. Within a certain subgroup the test can be predictive. But just like comparing across law schools, comparing across subgroups are more difficult. This is the part that he argues is unexplainable imo.


So your contention is that the historical data he's been using is biased.

In order for the bias to push down African American scores now, the historical data would have to measure African Americans that are, as a group, more intelligent than the African Americans who are currently being measured. So they show no difference on the tests, everything is good, then along comes the present generation who doesn't measure up, causing bias in the testing. My mind must be inflexible without caffeine, as this would be a strange thing to argue, and you must mean something else. How would you explain the situation in which the historical data was biased?


When they are testing experimental questions they are comparing the % correct in groups and as a whole to historical data. That is how they control for bias in a particular question. Maybe you should have a coke or something.

nycparalegal
Posts: 483
Joined: Sat Oct 17, 2009 11:26 am

Re: Interesting vid on urm lsats and admissions

Postby nycparalegal » Fri Feb 26, 2010 11:55 am

Most important fact that I learned from the video:

The 25% will become the 1st percentile for a lot of schools.

User avatar
ogman05
Posts: 266
Joined: Sun Apr 19, 2009 5:02 pm

Re: Interesting vid on urm lsats and admissions

Postby ogman05 » Fri Feb 26, 2010 11:59 am

dailygrind wrote:For the record, if there were a bias against African Americans while predictive validity was higher for them, it would just mean that a 160 for an African American was equivalent to a 164 or for a Caucasian, or some other number. It would be extremely easy for me to pick up in statistical analysis, and I definitely do not have a PhD specializing in this field.


Yes this is exactly the point. But they dont test for equivalency as in what a number for one group means to another group. The test only accounts for 16% after all. As in order to do this, as he mentions he would neeed a much larger test sample base. Its just not there except in HCBU's. and you cant base a whole test off one type of school. People would be more interested in T50 to T100 schools anyhow. If you can only apply it to some schools no one cares and statistical fail.

APimpNamedSlickback
Posts: 1126
Joined: Thu Feb 19, 2009 1:33 am

Re: Interesting vid on urm lsats and admissions

Postby APimpNamedSlickback » Fri Feb 26, 2010 12:00 pm

Mattalones wrote:I like the part about high high-scoring URMs are [strike]faught[/strike] fought over. ... Feelin' good! 8)


ftfy, but touche

User avatar
ogman05
Posts: 266
Joined: Sun Apr 19, 2009 5:02 pm

Re: Interesting vid on urm lsats and admissions

Postby ogman05 » Fri Feb 26, 2010 12:01 pm

Desert Fox wrote:
ogman05 wrote:
I understand all of what you are trying to say and was not trying to misquote him. I am only saying that according to "previous tests" there are no bias in the LSAT. But what are those tests based on. How do you control for bias? It is impossible unless you are using historical data which could in and of itself be biased. According to his statisticians there is no bias but how can they prove this I ask you? HISTORICAL data. That is how you run the regressions and other statistical calculators that they used. I took stats and upper level courses on this. If the data is biased your answer will be. End of story.

If 100 kids took a test and 20 were a certain group and all did worse. How do they test and control for that. They would either claim bias or compare it to a future test where 20 kids of that same group also did worse. There might not be any bias but there may. Within a certain subgroup the test can be predictive. But just like comparing across law schools, comparing across subgroups are more difficult. This is the part that he argues is unexplainable imo.


What he said was that the scoring differential exists on every single question. If it was a bias issues you'd likely see some questions be more biased that others, but they the scoring difference is the same on almost all questions. So he concluded that the LSAT is not a biased standardized test.

The difference in testing results continues into law school. They do as well as their LSAT suggests they will. That is another argument that the LSAT is an unbiased test.

Whatever is causing the difference is not likely a bias on the test, or on law school exams. Like he said, he doesn't know why it happens.


This I actually can agree with. This is the point that is the msot confusing but keep in mind thatthe scoring differential is the same on every question after they control for that diffential using historical tests. The interesting part is that it follows into schools. this could be for a number of reasons as well though that arent necesarrily causational from the test.

User avatar
dailygrind
Posts: 19639
Joined: Wed Oct 22, 2008 11:08 am

Re: Interesting vid on urm lsats and admissions

Postby dailygrind » Fri Feb 26, 2010 12:11 pm

ogman05 wrote:
dailygrind wrote:For the record, if there were a bias against African Americans while predictive validity was higher for them, it would just mean that a 160 for an African American was equivalent to a 164 or for a Caucasian, or some other number. It would be extremely easy for me to pick up in statistical analysis, and I definitely do not have a PhD specializing in this field.


Yes this is exactly the point. But they dont test for equivalency as in what a number for one group means to another group. The test only accounts for 16% after all. As in order to do this, as he mentions he would neeed a much larger test sample base. Its just not there except in HCBU's. and you cant base a whole test off one type of school. People would be more interested in T50 to T100 schools anyhow. If you can only apply it to some schools no one cares and statistical fail.


It is a piece of cake to test for what one number means for each group. All you have to do is run a regression on each group and see what the various coefficients line up for. Hell, you can just run a regression for the entire group, and then put in various dummy variables coding for race.

The population parameters of African Americans being different than it is for Caucasians is meaningless IMO. I think it's safe to assume that the population parameters for African Americans of the last, say 10 years, are the same parameters that African Americans have now. Pull from the same data, and the regression will still apply.

User avatar
D.Taggart
Posts: 2
Joined: Fri Oct 02, 2009 3:56 pm

Re: Interesting vid on urm lsats and admissions

Postby D.Taggart » Fri Feb 26, 2010 12:13 pm

dailygrind wrote:For the record, if there were a bias against African Americans while predictive validity was higher for them, it would just mean that a 160 for an African American was equivalent to a 164 or for a Caucasian, or some other number. It would be extremely easy for me to pick up in statistical analysis, and I definitely do not have a PhD specializing in this field.

But what would a Caucasian 164 mean? If you took your equivalence and said a Caucasian 164 was the same as a African American 160, it makes the Caucasian score sound just as predictive as the African American score. What I got from what he said was that (for example) a Caucasian 160 would range anywhere from African American 156 to a 164, in terms of law school performance.

User avatar
dailygrind
Posts: 19639
Joined: Wed Oct 22, 2008 11:08 am

Re: Interesting vid on urm lsats and admissions

Postby dailygrind » Fri Feb 26, 2010 12:20 pm

D.Taggart wrote:
dailygrind wrote:For the record, if there were a bias against African Americans while predictive validity was higher for them, it would just mean that a 160 for an African American was equivalent to a 164 or for a Caucasian, or some other number. It would be extremely easy for me to pick up in statistical analysis, and I definitely do not have a PhD specializing in this field.

But what would a Caucasian 164 mean? If you took your equivalence and said a Caucasian 164 was the same as a African American 160, it makes the Caucasian score sound just as predictive as the African American score. What I got from what he said was that (for example) a Caucasian 160 would range anywhere from African American 156 to a 164, in terms of law school performance.


Without having actual numbers in front of me, I can only make comparisons based off of "greater validity," so this is not very precise, but I will try to illustrate what I'm saying. Keep in mind, I do not believe that this is necessarily the case. I would imagine that the professor would have mentioned that this were the case in a 42 minute lecture if it were occurring. But if he didnt...

Let's say we accept my numbers. In that case, a 160 African American score would have a corresponding law school GPA, let's say 3.3, with a certain standard deviation, let's say 0.2. We expect the 160 African American to get a 3.3 in LS, and 96% of the time, we expect them to get between a 2.9 and a 3.7.

For the Caucasian, that score is a 164. Variance is higher for them because predictive validity is less. So we expect a 3.3, but the standard deviation is say 0.25. We expect that 96% of the time, the Caucasian with a 164 would get between 2.8 and 3.8. You see what I'm saying?
Last edited by dailygrind on Fri Feb 26, 2010 12:34 pm, edited 1 time in total.

User avatar
ogman05
Posts: 266
Joined: Sun Apr 19, 2009 5:02 pm

Re: Interesting vid on urm lsats and admissions

Postby ogman05 » Fri Feb 26, 2010 12:23 pm

dailygrind wrote:
ogman05 wrote:
dailygrind wrote:For the record, if there were a bias against African Americans while predictive validity was higher for them, it would just mean that a 160 for an African American was equivalent to a 164 or for a Caucasian, or some other number. It would be extremely easy for me to pick up in statistical analysis, and I definitely do not have a PhD specializing in this field.


Yes this is exactly the point. But they dont test for equivalency as in what a number for one group means to another group. The test only accounts for 16% after all. As in order to do this, as he mentions he would neeed a much larger test sample base. Its just not there except in HCBU's. and you cant base a whole test off one type of school. People would be more interested in T50 to T100 schools anyhow. If you can only apply it to some schools no one cares and statistical fail.


It is a piece of cake to test for what one number means for each group. All you have to do is run a regression on each group and see what the various coefficients line up for. Hell, you can just run a regression for the entire group, and then put in various dummy variables coding for race.

The population parameters of African Americans being different than it is for Caucasians is meaningless IMO. I think it's safe to assume that the population parameters for African Americans of the last, say 10 years, are the same parameters that African Americans have now. Pull from the same data, and the regression will still apply.


If you think this is such a pc of cake why dont you go show all of lsac and their statisticians how it is done and make their regression. Then you will have the nobel prize. Have you ever made your own regression before? I'm gonna go with no based on your answer that it is cake. People get phd's on this stuf and all sorts of degrees and still have difficulty. It is not that simple to end up with variables that matter and have not only correlation but causation as well.

In response to whether the same :"parameters" or data exist for AA's throughout the years I couldnt tell you. I have no data on that. I know that the acheivement gap has been closing though so I question that.

User avatar
dailygrind
Posts: 19639
Joined: Wed Oct 22, 2008 11:08 am

Re: Interesting vid on urm lsats and admissions

Postby dailygrind » Fri Feb 26, 2010 12:27 pm

Brosef, I'm getting my masters in Econ and I've done my fair share of regressions. I'm sure that they've done what I'm talking about, and I'm guessing that there's no difference.

User avatar
MC Southstar
Posts: 1238
Joined: Sat Sep 26, 2009 3:27 pm

Re: Interesting vid on urm lsats and admissions

Postby MC Southstar » Fri Feb 26, 2010 12:29 pm

Even I'm afraid to argue about math with an Asian.

nycparalegal
Posts: 483
Joined: Sat Oct 17, 2009 11:26 am

Re: Interesting vid on urm lsats and admissions

Postby nycparalegal » Fri Feb 26, 2010 12:31 pm

The most interesting question raised by the video is the following: why are African Americans mis-applying to law school?

User avatar
ogman05
Posts: 266
Joined: Sun Apr 19, 2009 5:02 pm

Re: Interesting vid on urm lsats and admissions

Postby ogman05 » Fri Feb 26, 2010 12:32 pm

dailygrind wrote:
D.Taggart wrote:
dailygrind wrote:For the record, if there were a bias against African Americans while predictive validity was higher for them, it would just mean that a 160 for an African American was equivalent to a 164 or for a Caucasian, or some other number. It would be extremely easy for me to pick up in statistical analysis, and I definitely do not have a PhD specializing in this field.

But what would a Caucasian 164 mean? If you took your equivalence and said a Caucasian 164 was the same as a African American 160, it makes the Caucasian score sound just as predictive as the African American score. What I got from what he said was that (for example) a Caucasian 160 would range anywhere from African American 156 to a 164, in terms of law school performance.


Without having actual numbers in front of me, I can only make comparisons based off of "greater validity," so this is not very precise, but I will try to illustrate what I'm saying. Keep in mind, I do not believe that this is necessarily the case. I would imagine that the professor would have mentioned that this were the case in a 42 minute lecture if it were occurring. But if he didnt...

Let's say we accept my numbers. In that case, a 160 African American score would have a corresponding law school GPA, let's say 3.3, with a certain standard deviation, let's say 0.2. We expect the 160 African American to get a 3.3 in LS, and 96% of the time, we expect them to get between a 2.9 and a 3.7.

For the Caucasian, that score is a 164. Variance is higher for them because predictive validity is less. So we expect a 3.3, but the standard deviation is say 2.5. We expect that 96% of the time, the Caucasian with a 164 would get between 2.8 and 3.8. You see what I'm saying?


What is "greater validity"? Just because someone has a 42 minute lecture they are not going to mention every single thing that has been researched for over half a century. that whole premise you have been using is complete fail. Second you cant complare across law schools which is why your "equivalency" argument doesnt work. If they did that then there would be no point in using the lsat for admissions criteria at all. Maybe, just maybe in peer schools but not all schools becuase of a thing called a curve and no std grading system that applies universally across all schools taking ALL law school students into consideration. Your third paragraph makes not sense to me and soundsd like you are just shooting out statistics terms.
Last edited by ogman05 on Fri Feb 26, 2010 12:34 pm, edited 1 time in total.




Return to “Choosing a Law School”

Who is online

Users browsing this forum: kalvano, MSNbot Media and 5 guests