Bombing Intro to Stats

[Umlaut]

So due to my choice of a Math minor I was forced to hear Intro to Stats at the Math department where the course in this form was held for the first time by a prof who overestimated pretty much everybody in there.

 

The material covered about three Semester of Stats/Metrics as it would be done in an Econ Stats lecture and did it in a much mire rigorous style. Protests that the material covered was too much did not elicit much of a response.

 

I do not know what my final grade will be, but I do know it will not be good (at all), as will most other peoples'.

 

In addition to mistakes that were inevitable, I also added one or two which I really shouldn't have, possibly because the onsetting panic unsettled me sufficiently. You can't resit passed exams where I am from and I am now scared shitless that a probably dismal grade in something falsely called "Intro to Stats" could destroy an otherwise decent profile.

 

Would one terrible grade be overlooked if I now try to remedy it by doing well in i.e. Metrics for Math students?

[Food4Thought]

If you get information on how well you did relative to your peers, you could potentially have one of your letter writers put the grade into context.

 

A terrible grade won't be overlooked. But it helps if your grade was less terrible than most.

[Blanket]

There's a lot of time left in the semester. You could recover or, more likely, the grade distribution will be returned to normal by way of the ever merciful curve. Good luck.

[Umlaut]

This is an exam from the last semester.

 

And the entire grade was 100% this exam.

[Umlaut]

For reference, would you consider this to be "overkill" for course with two lectures and one TA session a week?

I'm trying to convice myself that it was in order to give myself some kind of excuse, but would really appreciate your honest opinion.

For me this was the first time that I really felt I couldn't keep up with the amount of stuff that was being thrown at me.

 

Everything was introduced via theorem-proof with problem sets consisting to 80% of either proofs or tasks in R.

Noone of us had had anything to do with statistics since high school. Introduction to probability theory was held in parallel.

 

- Multivariate descriptive statistics and Multivariate Linear Modells/Regression

- (Pre-)Hilbert Spaces (Convergence, Cauchy-Schwarz, Projections etc.)

- Principal Components Analysis

- Cluster Analysis (including introduction of string metrics, i.e. M/S-coefficients, Mahalanobis-norm, etc.)

- Factor Analysis via one of the ugliest proofs I've ever seen

- Proof of Central Limit Theorem for heteroscedastic variables via Huber (probably the ugliest proof I have seen so far)

- Multivariate Normal Distribution (properties and proofs of properties)

- Estimators: BLUE, Method of Moments, M-Estimators (Maronna), ML-Estimators including proofs of Bias, MSE and asymptotic convergence

- Robust Statistics: Formal definition of Breaking Point/Influence curves

- Tests: (two-sample-)t-Test, Chi-Squared, etc. including Behrens-Fischer-Problem

- Neyman-Pearson-Lemma/Estimators, predominantly proof of their properties

- Unbiased Estimators: Fischer Neyman-factorisation theorem, Rao-Blackwell, Lehmann-Scheffé (including proofs)

- Bayesian approach/Bayes-Risk

- Jackknife

- Parametric and non-parametric Bootstrap

- Discriminant function analysis based on Kernel density Estimators

 

Should anyone want the notes to this, I guess I could provide them if you send me a PN - though they wouldn't be in English and are about half as extensive and long as they ought to be if they were to provide enough intuition and explanation behind the proofs.

Edited February 15, 2015 by Umlaut

[sevet]

Are those all the topics covered in that one class?

[Food4Thought]

All the more reason to have someone comment on the ridiculousness of the class. If this is truly the level of the class, which is hard to believe, then no one in their right mind would would hold a less than stellar grade against you.

[Umlaut]

This was the material covered in 28 lectures held over 14 weeks... (**** me, right?)

[Econhead]

This was the material covered in 28 lectures held over 14 weeks... (**** me, right?)

 

It's quite extensive, but I have had classes that covered a similar amount of material. There is usually a method to the madness, and by that I mean that if the professor covers that much material there are usually only 1 or 2 students that are separate from the pack in terms of grades (at the top end), however even these students didn't do so well to as to earn an A pre-curve. All other students that earned A's typically have significantly lower scores, and there is a heavy curve applied. If the average (after the curve) was far below a B or B-, the professor was probably not adjusting enough for the difficulty of the course. Of course, if you end up at the top end of the grading scale, you typically think that it was a lot of work, but feel that it was (moderately) unfair that so many students earned A's when they didn't do nearly as well as you. I guess it's just one of those things.

 

What level of university are you at, Umlaut? At my undergrad (unranked), we covered material down to the nitty gritty of really understanding the proofs, but we covered it more slowly. What we did in a prob course and a stat course ended up being covered in 1 semester at my current university (master's coursework, T70). Even though the same material was (broadly) covered, and in some ways we covered more, it was not nearly as intense, and there was a moderate curve. At my undergrad, there was absolutely no curve (beyond maybe 1-3%). I learned significantly more at my undergrad institution, but I was thankful to have been exposed to more material at my graduate institution.

[fakeo]

It's quite extensive, but I have had classes that covered a similar amount of material. There is usually a method to the madness, and by that I mean that if the professor covers that much material there are usually only 1 or 2 students that are separate from the pack in terms of grades (at the top end), however even these students didn't do so well to as to earn an A pre-curve. All other students that earned A's typically have significantly lower scores, and there is a heavy curve applied. If the average (after the curve) was far below a B or B-, the professor was probably not adjusting enough for the difficulty of the course. Of course, if you end up at the top end of the grading scale, you typically think that it was a lot of work, but feel that it was (moderately) unfair that so many students earned A's when they didn't do nearly as well as you. I guess it's just one of those things.

 

I'm guessing @Umlaut is in Europe, where grading on a curve is not common at all. I mean I've seen/heard of several courses where the best grade was (the local equivalent of) a D.

[Econhead]

I'm guessing @Umlaut is in Europe, where grading on a curve is not common at all. I mean I've seen/heard of several courses where the best grade was (the local equivalent of) a D.

 

I accept that this happens, but I don't understand what the point of doing this is. It could grossly affect an individual's future. Further, even if you are going to be a professor that is going to grade very difficult, or test their students over so much information that it is extremely difficult to get a good grade without a curve, I don't understand why you wouldn't still try to break up the student's grades into some sort of "normal grading scale," based on how you think they are doing. For example, suppose there is a relatively normal distribution of grades, but only 0-70% is used. I don't understand why you wouldn't just divide the scale such that (perhaps) you believe the kids who scored the following understood the material for the following grades:

 

A 55-70%

B 40-55

C 32-40

D 25-32

 

It just makes no sense to me what purpose assigning the top student a D serves.

 

I understand that what I have suggested above is effectively "grading on the curve," but to me it's much different than simply saying that you are going to raise everyone's grade by, say, 30%. You're looking at the student's work and trying to actually separate what you think their grade should be based on what they have demonstrated. Although it might seem odd to break letter grade's up by more or less than 10% (varying across letters), I have seen this done before, and makes sense to me when trying to really assess how much a student has learned.

[fakeo]

I accept that this happens, but I don't understand what the point of doing this is. It could grossly affect an individual's future.

 

There isn't a point to it, it's just a different culture. In Europe GPA is not as important of a metric as in the US. So in principle, your future won't be affected by it so much (at least if you stay in your home country or if foreign adcoms are familiar with your school's grading culture, which is almost always the case).

 

I think Umlaut should be alright. One grade won't break a profile. Good LORs and good grades in other courses can make up for one bad grade. And if such "grade deflation" happens often in his/her country, then the adcoms will likely be aware of it.

[MR14]

I accept that this happens, but I don't understand what the point of doing this is. It could grossly affect an individual's future. Further, even if you are going to be a professor that is going to grade very difficult, or test their students over so much information that it is extremely difficult to get a good grade without a curve, I don't understand why you wouldn't still try to break up the student's grades into some sort of "normal grading scale,"

 

 

An advantage of not grading on a curve is that grades across years are a bit more comparable. If your year is especially good, a curved grade would not reflect your performance compared to students from other years. Although I agree that giving all students D's might be a bit too much, I would note that employers (locally) are familiar with grading practices, and adapt their expectations accordingly.

[Umlaut]

Well results are out and I scored (the equivalent of) a D+.

The best grade I've heard of so far is a C-.

So much for my (anyway ambitious) goal of graduating with distinction.

 

Depending on how things work out, I think I wouldn't do the Math Minor/Major again.

Despite the fact that you do learn a lot, the grading and the fact that you're competing with people without the burden of an additional 10-15 hours of work is just too much.

Also, I think people aren't too focussed on the lacking Math when comparing European graduates.

[fantinity]

I actually had a kind of similar situation: I had C+ for math course 'optimization methods' (sounds bad enough), and lost distinction and scholarship (there is no tuition in Russia) just because of an a**hole professor who had a bad day (verbal exam). It all worked out in the end, but it's worth mentioning that I got MSc in Economics afterwards (same school), and we had a Math for Economists course there (A-). None of this was mentioned in SOP or LORs, I figured it was a long time ago, and math major kind of makes up for one bad grade.

[Econhead]

I actually had a kind of similar situation: I had C+ for math course 'optimization methods' (sounds bad enough), and lost distinction and scholarship (there is no tuition in Russia) just because of an a**hole professor who had a bad day (verbal exam). It all worked out in the end, but it's worth mentioning that I got MSc in Economics afterwards (same school), and we had a Math for Economists course there (A-). None of this was mentioned in SOP or LORs, I figured it was a long time ago, and math major kind of makes up for one bad grade.

 

Judging by your results, you're clearly wrong. Only Harvard or MIT would have been acceptable.

 

:stupid: