posted
One of the judges I'm applying for has asked for additional information, including class rank. He specifically states, "Even if your law school does not compute standings, give it your best shot."
Here's my best shot:
quote:My class has approximately 360 students. The mean GPA is 3.3; a GPA of 3.48 represents the top 25% of the class. Assuming a normal distribution, the standard deviation is 0.2669. Given a normal distribution with mean 3.3 and standard deviation .2269, my GPA of 3.873 places me in the top 1.59% of the class, which places me at approximately 6 out of 360.
There are two major problems with the estimate. First, there's no way of knowing how close the distribution of GPAs are to a normal distribution. Second, it's very possible that even if it is close to a normal distribution, the tails are fatter. Even a small deviation could change my class rank by a significant amount.
Given that, this is still my best guess. My questions are:
1.) Is the quoted explanation accurate and understandable to someone with basic statistical knowledge?
2.) How much, if any, do I need to mention the possible issues with the estimate?
posted
I think a disclaimer that small changes in GPA could produce large swings in rank at the extremes of the curve (maybe with a subtle hint about which extreme you find yourself in ) would be enough. I didn't verify your calculations but they make sense to me.
Posts: 3846 | Registered: Apr 2004
| IP: Logged |
Have you tried asking your school's records clerk or career counselors for a rough estimate?
My school also does not provide class rankings. However, if a student made it clear that he really needs a rough estimate of his class rank in order to secure a job, the career counselors are usually pretty cooperative.
Posts: 4116 | Registered: Apr 2002
| IP: Logged |
posted
I'll try that, Beren, but I'm skeptical. They've been pretty adamant, and they've stated they don't want to dilute the blanket refusal with exceptions.
Assuming I can't get anything from career services, how about adding this sentence at the end of the description:
"The possibility that the distribution of GPAs is not a normal distribution and the greater possibility of the GPA distribution diverging from the normal distribution at the high end of the curve may lead to small errors in this calculation."
Posts: 26071 | Registered: Oct 2003
| IP: Logged |
posted
I would emphasize that the assumption of a normal distribution is a weak link in the calculation.
You could also put bounds on your class rank, obviously you're in the top 25%, and you could improve that bound.
Given the mean and the top 25% number, you can probably estimate the deviation better than you would get from using a normal distribution assumption, which would give more confidence in your rank estimate.
I have to go to work but I'll try to get back with you on that deviation estimate later.
Posts: 6316 | Registered: Jun 2003
| IP: Logged |
posted
I would just say I'm somewhere in the top 10 and leave it at that. But I'm an engineer, so I don't like to get too specific about things.
Posts: 4625 | Registered: Jul 2002
| IP: Logged |
posted
It's not a possibility, it's a certainty that the distribution is not normal, which is why it's a weak assumption, but without more info from the registrar, perhaps an unavoidable one.
Posts: 6316 | Registered: Jun 2003
| IP: Logged |
posted
"They've been pretty adamant, and they've stated they don't want to dilute the blanket refusal with exceptions."
That's our school's official stance as well. But you never know until you try.
I bugged my counselor a couple of times until she finally gave in. I emphasized the fact that the employer specifically askef for a class ranking, and if I'm the only applicant who could not provide that information, I would be at an disadvantage.
Posts: 4116 | Registered: Apr 2002
| IP: Logged |
quote:I have to go to work but I'll try to get back with you on that deviation estimate later.
That would be great. My statistics abilities don't go much beyond T-distributions, chi-square tests, and linear regression. All of those require a sample to work off.
But don't put yourself out. I think top 10 is very safe. 5% is almost certain, and 10% is a dead lock.
Part of the reason I'm including this description is to give him an idea of how I operate. It would drive some people crazy that I include the detail; others would like it. It's best that he know now what I'm like. Posts: 26071 | Registered: Oct 2003
| IP: Logged |
posted
Whether it's wholly accurate or not, I think it's impressive the way you estimated it, and I would definitely include it.
It reminds me of the sample problems I often got asked in interviews (Electrical engineer) - questions like: How many quarters stacked one on top of another would it take to reach the peak of the empire state building?
They didn't care about your answer, they just wanted to see how you approached the problem.
Posts: 5462 | Registered: Apr 2005
| IP: Logged |
posted
Four. Obviously. Four quarters of the Empire State building stacked on top of each other exactly reaches the peak, if you slice the quarters horizontally.
Posts: 1466 | Registered: Jan 2003
| IP: Logged |
posted
The only problem I could see with your distribution is the assumption that the distribution is normal. This could change your results greatly, you may be alot higher or lower than 6! Especially with something like class distribution. Otherwise, it is understandable from a statistical view.
Posts: 457 | Registered: Jun 2005
| IP: Logged |
posted
You should say that your rank is 6th, plus or minus 2 at the 95% confidence level.
Actually, I would think that if you are safe saying you're in the top 5%, you should do that and just add a note to say that you had to work it out using the mean and standard deviation since your school doesn't provide rankings. I wouldn't want to work for a judge who didn't understand the concept of mean and standard deviation anyway. So if this person doesn't buy it or doesn't understand the concept, you should reject the job offer as beneath you.
Posts: 22497 | Registered: Sep 2000
| IP: Logged |
posted
Based on a policy at Duke's law school which enforces the basics of a certain distribution in classes over 40 students, supposedly because of community expectations, I suspect the distribution is bimodal, with a small hump from 3.7-4.0 and a larger one around 3.0-3.3.
I think I can make a pretty decent economic argument for why this might be so, too, based on consumer equilibrium and some simple assumptions.
Posts: 15770 | Registered: Dec 2001
| IP: Logged |
posted
fugu, there is no distribution policy. I've only looked at the distributions in my classes, and all of them were unimodal: more As than A+s, more A-s than As, more B+s than A-s, fewer Bs than B+s, etc. Some had fatter tails than a true normal would, but all were basically symmetrical. Also, grades may only be given in increments of 4.3, 4, 3.7, 3.3, 3, 2.7, etc. Nothing in between may be given, although of course GPAs fill in the gaps.
This doesn't mean that there isn't a bimodal distribution of GPAs, of course. It's still very possible that someone who gets an A is more likely to get another one. But from a sample of 2 years' worth of classes, grades in individual classes are not bimodal.
Unless Morbo can give a better estimate of the distribution, I'll probably go with a strongly worded caution about likely error.
He said "best shot," and all the info I have is included in that analysis.
Posts: 26071 | Registered: Oct 2003
| IP: Logged |
posted
Even without the distribution policy, I'd be oddly lured to the bimodal.
You mention the 2 years worth of classes though; do you have any further info on them, even slightly? Because if you did, you could come up with a darn good estimate (since they each constitute random samples of the population, basically).
Posts: 15770 | Registered: Dec 2001
| IP: Logged |
posted
I can look up the breakdown by grade for each class. But without knowing which grades go with which person, I won't have a random sample of GPA, just of grades assigned.
Posts: 26071 | Registered: Oct 2003
| IP: Logged |
posted
It won't give you GPA, but it will give you the range. So that'll give you a little more accurate reading regardless of which distribution you go with.
Posts: 5462 | Registered: Apr 2005
| IP: Logged |
posted
It doesn't matter if you know which grades go with which person for our purposes, the correspondence between grades and GPA points is enough. Then each class can be treated as a random sample and the overall distribution estimated.
Posts: 15770 | Registered: Dec 2001
| IP: Logged |
Printer-friendly view of this topic
![[Smile]](smile.gif)
![[Wink]](wink.gif)
