Doing Better at a "Worse" School Forum

[Volake]

you're taking a barely statistically significant apple and comparing it to an orange. it's not just minimally decisive, it's quite likely misleading.

So you're conceding that it could be decisive? Misleading? Why do you think that a reasonable applicant couldn't properly discount the value of the factor?

[AllTheLawz]

And what we are telling you is that this is a dumb factor to consider because it is not predictable at the individual level and, once you discount for that fact, any effect is insignificant and likely within the reasonable margin for error. You might as well factor in your comfort with testing under timed pressure and then weigh the number of profs offering take-home exams vs in-class exams. Or the average temperature in December to account for the likelihood of you getting sick during first semester exams.

Yes, there are people that UGPA and LSAT correspond to law school test performance better than others. But I don't see how that makes it "dumb." Why would a law student prefer 10% boost to .5 probability of 20%?

EDIT: And in response to the person who says that the original poster's question was answered. The answer the majority seems to be giving assumes that little = nothing .

You just completely misinterpreted and its obvious you have no clue what you are talking about. Put simply: The equation you are actually talking about is one with a set of variables with certain magnitudes that effect an outcome. This isn't a case where simple probabilities would be helpful. In this case, the magnitude of the effect of the "small pond" variable would be so low that even if it were somehow statistically significant, any effect it could possibly have would probably be dwarfed by the margin for error. That is to say, it has absolutely zero predictive value.

[Clearly]

[Volake]

Reply with quote
Volake wrote:
Quote:
And what we are telling you is that this is a dumb factor to consider because it is not predictable at the individual level and, once you discount for that fact, any effect is insignificant and likely within the reasonable margin for error. You might as well factor in your comfort with testing under timed pressure and then weigh the number of profs offering take-home exams vs in-class exams. Or the average temperature in December to account for the likelihood of you getting sick during first semester exams.


Yes, there are people that UGPA and LSAT correspond to law school test performance better than others. But I don't see how that makes it "dumb." Why would a law student prefer 10% boost to .5 probability of 20%?

EDIT: And in response to the person who says that the original poster's question was answered. The answer the majority seems to be giving assumes that little = nothing .


You just completely misinterpreted and its obvious you have no clue what you are talking about. Put simply: The equation you are actually talking about is one with a set of variables with certain magnitudes that effect an outcome. This isn't a case where simple probabilities would be helpful. In this case, the magnitude of the effect of the "small pond" variable would be so low that even if it were somehow statistically significant, any effect it could possibly have would probably be dwarfed by the margin for error. That is to say, it has absolutely zero predictive value.

Forgive me for failing to see the obvious. How low, by your thinking, is low? 1%, 5%, 10%? And I find it difficult to believe that you think that between a random large set of T1 students and a random large set of students at T14 school there would be no statistically significant difference between their grades in a class taught by the same professor.

Now, if you concede that there would be, but that we cannot say to whom of the t14 that advantage would attach, I contend that this doesn't matter. The advantage of being an outsize recipient of the small pond factor would just about balance the disadvantage of not sharing in the advantage, ex ante.

[patogordo]

they aren't being taught by the same professor. they aren't even being graded on the same scale. the correlation between lsat and 1L gpa is miniscule, and barely statistically significant, and that's for students at the same school taking the same classes on the same curve with roughly the same professors.

[cron1834]

Volake = "Add to foe list"

Are you sure you want to carry out this operation?

Yes.

[Pneumonia]

How the hell did this thread even get to 4 pages?

The OP has had his answer multiple times from the look of things. Special snowflakes don't melt in the heat of the words on a computer screen guys. It takes the harsh sunshine of reality. Stop posting and this situation will resolve itself on its own.

preach

[PepperJack]

This question is answered on these forums like once a week at least. There is some correlation between LSAT score and grades and your odds of doing better might be slightly higher at a lower ranked school but the employment prospects drop much faster than any advantage gained by going to a lower ranked school. If your choices were going to a state flagship for cheap vs. t-14 at sticker maybe this would come into play but no matter what you should understand that even being above a schools 75th percentile in GPA and LSAT score might leave you below median if law exams just aren't your skill set.

Just want to say as this issue goes on and on full of speculation and guesses, and generalizations from anecdotal stories of someone's friend's dog's hooker that there is substantial data on this issue. This data suggests exactly what the post above says. There is definitely some correlation. However, it's only good at predicting odds in extremes. As extremes are 10 points or more, it's likely to change to a place where the improved rank still isn't good enough. It'd only have value if one were choosing between Yale and a full ride to Georgetown with a 180 (statistically, they'd be more likely [no means a lock] to do well at Georgetown]). Financially, there would be an argument for Georgetown. There'd also be a counterargument. However, I wouldn't see an argument ever unless the lower school is a full ride offer being the likelihood of "winning but losing".

This can't be tested for, but given the confidence and work ethic of people who generally do well on the exams, it seems those who would seek out less competitive environments may also be less likely to perform. No data, just a personal observation of zero empirical use.

[Volake]

Just want to say as this issue goes on and on full of speculation and guesses, and generalizations from anecdotal stories of someone's friend's dog's hooker that there is substantial data on this issue. This data suggests exactly what the post above says. There is definitely some correlation. However, it's only good at predicting odds in extremes. As extremes are 10 points or more, it's likely to change to a place where the improved rank still isn't good enough. It'd only have value if one were choosing between Yale and a full ride to Georgetown with a 180 (statistically, they'd be more likely [no means a lock] to do well at Georgetown]). Financially, there would be an argument for Georgetown. There'd also be a counterargument. However, I wouldn't see an argument ever unless the lower school is a full ride offer being the likelihood of "winning but losing".

This can't be tested for, but given the confidence and work ethic of people who generally do well on the exams, it seems those who would seek out less competitive environments may also be less likely to perform. No data, just a personal observation of zero empirical use.

The question now becomes what to do in the absence of the data we'd like. What is your Bayesian prior, as it were, with regard to the matter? The data from transfer students in various threads, LSAC, and common sense suggests that people with better LSAT scores and ugpas tend to do better on law school exams. My Bayesian prior on this would be that between, for instance, a random t14 and a random t1, the difference would not be less than 5% in the aggregate, which is at least worth considering.

[Clearly]

Just want to say as this issue goes on and on full of speculation and guesses, and generalizations from anecdotal stories of someone's friend's dog's hooker that there is substantial data on this issue. This data suggests exactly what the post above says. There is definitely some correlation. However, it's only good at predicting odds in extremes. As extremes are 10 points or more, it's likely to change to a place where the improved rank still isn't good enough. It'd only have value if one were choosing between Yale and a full ride to Georgetown with a 180 (statistically, they'd be more likely [no means a lock] to do well at Georgetown]). Financially, there would be an argument for Georgetown. There'd also be a counterargument. However, I wouldn't see an argument ever unless the lower school is a full ride offer being the likelihood of "winning but losing".

This can't be tested for, but given the confidence and work ethic of people who generally do well on the exams, it seems those who would seek out less competitive environments may also be less likely to perform. No data, just a personal observation of zero empirical use.

The question now becomes what to do in the absence of the data we'd like. What is your Bayesian prior, as it were, with regard to the matter? The data from transfer students in various threads, LSAC, and common sense suggests that people with better LSAT scores and ugpas tend to do better on law school exams. My Bayesian prior on this would be that between, for instance, a random t14 and a random t1, the difference would not be less than 5% in the aggregate, which is at least worth considering.

I'm just dying to know more about you personally. Did you attend law school? where?

[PepperJack]

No, there is actual data. Like extensive data publicly available in 100s of law review articles. Granted, most of it is from pre-2000. Also, some of it is used to look at the impact of race based admission preferences. However, the actual data being utilized doesn't break up students up by race. It breaks them up by LSAT score, and looks at the class rank with the LSAT score. But nobody at LSAC would argue that they'd be confident predicting a 168 will outperform a 165. Vegas might consider it 2:1 odds that a 168 would do better than a 158. The only place this decision should matter to someone is if you get a full ride to a worse school. I buy the logic of taking the full ride over paying sticker, because you have a decent chance of the class rank causing it to balance out while walking away with no debt. Hypothetically, if you dominated a t-20, it's not unreasonable to presume you'd have dominated a TT as well. If TT is free then you just produced the best possible outcome for yourself of the whole equation (job + no debt). It's not an unreasonable equation.

[patogordo]

they aren't being taught by the same professor. they aren't even being graded on the same scale. the correlation between lsat and 1L gpa is miniscule, and barely statistically significant, and that's for students at the same school taking the same classes on the same curve with roughly the same professors.

[PepperJack]

Just want to say as this issue goes on and on full of speculation and guesses, and generalizations from anecdotal stories of someone's friend's dog's hooker that there is substantial data on this issue. This data suggests exactly what the post above says. There is definitely some correlation. However, it's only good at predicting odds in extremes. As extremes are 10 points or more, it's likely to change to a place where the improved rank still isn't good enough. It'd only have value if one were choosing between Yale and a full ride to Georgetown with a 180 (statistically, they'd be more likely [no means a lock] to do well at Georgetown]). Financially, there would be an argument for Georgetown. There'd also be a counterargument. However, I wouldn't see an argument ever unless the lower school is a full ride offer being the likelihood of "winning but losing".

This can't be tested for, but given the confidence and work ethic of people who generally do well on the exams, it seems those who would seek out less competitive environments may also be less likely to perform. No data, just a personal observation of zero empirical use.

The question now becomes what to do in the absence of the data we'd like. What is your Bayesian prior, as it were, with regard to the matter? The data from transfer students in various threads, LSAC, and common sense suggests that people with better LSAT scores and ugpas tend to do better on law school exams. My Bayesian prior on this would be that between, for instance, a random t14 and a random t1, the difference would not be less than 5% in the aggregate, which is at least worth considering.

More qualified people assess it. Another caveat with the LSAT data that could string it into working in either direction are the special snowflakes who get into law school with much worse scores. Are they underperforming their scores because they're special snowflakes, or are they overperforming them? The prediction is only valuable in extreme discrepancies in scores. Hence, you are really comparing the median student to a special snowflake, not a regular person with that score. So really, there's no data at all.

[Volake]

they aren't being taught by the same professor. they aren't even being graded on the same scale. the correlation between lsat and 1L gpa is miniscule, and barely statistically significant, and that's for students at the same school taking the same classes on the same curve with roughly the same professors.

1) "They aren't being taught by the same professor being graded on the same scale."

Here, we're comparing how a student would fare in two counterfactuals: student taking a 1L test (Torts for instance) at a higher ranked school after a semester vs. student taking the test at a lower-ranked school. We can compare where statistically the student will be at at when her lsat/ugpa stats are among the lower group of scores and where statistically they'll be ranked among the higher group of scores.

So, what we'd like to be able to test would be repeated series of a student with different input combinations attending semesters of law school and see where, on average they place. Absent this information, we can only use the information we do have (from LSAC, limited transfer data) in conjunction with our intuitions to think about how large an effect the "small pond" would have.

2) "barely statistically significant"

http://www.lsac.org/docs/default-source ... rmance.pdf

LSAC says that the two input metrics, taken together, have a .48 correlation with first-year performance. This conforms with anecdotal data from TLS suggesting a slight drop on average in the ranking of transfer students.This is not the highest correlation especially when you consider that you're not looking at extremes. But it suggests that you will perform modestly better at a significantly lower-ranked school (i.e. t14 vs. t1). I don't know what precisely how it would translate, on average, but my intuition suggests that it's not negligible, and could factor (among many other things) into a difficult choice.

I am not arguing that it's a huge factor, I'm saying that it reasonably could be determinitive (at least worth considering) for people who have tough choices between schools that are significantly apart in the input score bands of their students ($$$$ T1 v. sticker T14)

I'm just dying to know more about you personally. Did you attend law school? where?

2L at Chicago

[patogordo]

I'm just dying to know more about you personally. Did you attend law school? where?

2L at Chicago

flame

[brotherdarkness]

.

[woosah]

It would be prudent for the applicant to consider the small pond factor in the entirety of their analysis, while acknowledging that it's only a small advantage.

I completely agree with this. Furthermore, I don't think the data is conclusive enough to relegate it to "completely dismiss-able" as some posters are suggesting. The only data I have on it is what I included in my OP. I've seen many posters claim that "the data shows that it is wildly negligible," but personally I haven't seen this data.

[Clearly]

The Chicago thing makes sense.

[rickgrimes69]

I'm just dying to know more about you personally. Did you attend law school? where?

2L at Chicago

flame

it was well played up until then