Hey, Math People, I Have a Couple of Data Points, Help?!? Forum

[KMaine]

OK, so I am interested to know where I stand in terms of class rank. But our school is pretty tight-lipped about the curve. But here is what I know:

Mean = 3.35 (I know this is different from median, but let's suppose they are the same
30th%ile = 3.47
10th%ile = 3.66
5th%ile = 3.78

Is there any way of telling about where 15th, 20th, 25th, 35th, 40th etc. are?

[Other25BeforeYou]

How do you know the 30th %ile?

Also, I never made it into school this week at the same time you did, so I am giving you a virtual high five. HIGH FIVE!

[09042014]

15% ~= 3.609
20% ~= 3.55
35% ~= 3.44
40% ~= 3.40

These aren't exact because it isn't a truly normalized scale. But it's a fair approximation.

[KMaine]

Other 25 - I will take the real thing next week.

DF - Thanks for the estimate!

[09042014]

Other 25 - I will take the real thing next week.

DF - Thanks for the estimate!

I used a .25 standard deviation for 15% and .24 for the rest.

[Dr. Strangelove]

If law schools are tight-lipped about their curves to mitigate stress levels of law students, it's a pretty dumb reason.
I second DF's approximation.

[09042014]

If law schools are tight-lipped about their curves to mitigate stress levels of law students, it's a pretty dumb reason.
I second DF's approximation.

Good law schools do it so employers can't make strict cut offs. Less good law schools usually tell students their rank exactly.

[thesealocust]

nm

[09042014]

These aren't exact because it isn't a truly normalized scale. But it's a fair approximation.

Actually, you'd be surprised

<---- has done these calcs for people assuming a normal distribution and had it confirmed when actual rankings came out in almost every instance

I know it isn't because he gave a couple data points and I calculated different variance when I used the different points. The standard deviation for the 5% worked out to .26, and the one for 30% worked out to be .23.

It's going to be pretty accurate.

[thesealocust]

nm

[Alyosha]

Sealocust (or DF),

Is this true even at the margins? If you calculated the top 5% assuming a normal distribution, and using last year's data for top 10% and median, has it generally been confirmed when rank came out?

Also for some reason the GPA required for top 10% appears to drop at my school in the 2nd and 3rd years. My school has a mean requirement (it's essentially the same requirement it was for 1L) but no curve requirement for upper level courses. Wouldn't upper year GPA's increase under this scenario? For example the nalp data for my school says top 10% at graduation was 3.69, while my school's published top 10% cutoff for last year's 1L class was 3.76. This seems strange.

Thanks in advance for any help (or unsubstantiated guesses)

[miamiman]

df, briefly, how are you calculating the SD?

[Tautology]

All of you are using the wrong percentile numbers. The top 5% is the 95th percentile. Other than that though, DFs numbers look pretty good from a quick glance.

[09042014]

df, briefly, how are you calculating the SD?

By solving the probability density function for a normalized curve, at the percents he gave me. If you have the median, and the rank at one spot, you can solve for the SD.

[miamiman]

df, briefly, how are you calculating the SD?

By solving the probability density function for a normalized curve, at the percents he gave me. If you have the median, and the rank at one spot, you can solve for the SD.

but we only have the mean.

[09042014]

df, briefly, how are you calculating the SD?

By solving the probability density function for a normalized curve, at the percents he gave me. If you have the median, and the rank at one spot, you can solve for the SD.

but we only have the mean.

Yea you are right, I didn't even read that the first time.

Technically you could set up a system of equations and solved for the two variables but that will take a lot more work.

[thesealocust]

nm

[miamiman]

better question: lawyers are, generally speaking, fucking math retarded. do they just eyeball it and say... "it's PROBABLY this" or do they have the dir. of recruiting crunch numbers

[Tautology]

df, briefly, how are you calculating the SD?

By solving the probability density function for a normalized curve, at the percents he gave me. If you have the median, and the rank at one spot, you can solve for the SD.

but we only have the mean.

Yea you are right, I didn't even read that the first time.

Technically you could set up a system of equations and solved for the two variables but that will take a lot more work.

In a normal distribution, which we're assuming, the mean is the same as the median.

[09042014]

By solving the probability density function for a normalized curve, at the percents he gave me. If you have the median, and the rank at one spot, you can solve for the SD.

but we only have the mean.

Yea you are right, I didn't even read that the first time.

Technically you could set up a system of equations and solved for the two variables but that will take a lot more work.

In a normal distribution, which we're assuming, the mean is the same as the median.

Ah, bingo. Don't blame me I barely got a B in Stat.

[miamiman]

I got an A+ in stat (at HYPS no less) and don't remember shit to save my life.

[thesealocust]

nm

[Tautology]

The thing I can never remember is which are type I and which are type II errors.

[miamiman]

better question: lawyers are, generally speaking, fucking math retarded. do they just eyeball it and say... "it's PROBABLY this" or do they have the dir. of recruiting crunch numbers

BAM! You're a hiring director. You get 100 applications for an interview. You can only pick 40 to interview. Do you:

A) Pick the best 40 resumes

B) Attempt to eyeball GPA cut offs and ensure that any resume with a GPA below your guess doesn't get further consideration

C) Have the director of recruiting crunch numbers, create a GPA cut off, and ensure that any resume with a GPA below the calculation doesn't get further consideration.

It's not fucking rocket science. If 160 people bid on Cravath, they'll pick 80 interesting people to interview. When it comes time to make callback decisions, they'll look at GPAs + interview notes, factor in expectations based on experience, and then call some people back.

Put simply: They have so many resumes that they have no need for calculating a percentile - they can just say "oh, this d00d has a GPA lower than 70 of my 80 resumes, probably not going to call him back". Percentiles are handy for us because we tend to have many fewer data points to work with.

I got an A+ in stat (at HYPS no less) and don't remember shit to save my life.

Thanks for that. I predict good things for your social life when you get to law school.

Dude, you love to hate but I don't see what I said as being pretentious/obnoxious/otherwise negative. THANKS THOUGH for your re-education of sorts.

[thesealocust]

nm