What university level statistics courses are considered advanced/hard? I've recently been looking for top-of-the-line statisticians in a recruiting process for our company. Myself, I'm a Physics Engineering major. I gather that great mathematical statisticians have studied a bit different courses, and much more in depth. 
When evaluating a candidate, are courses a good indicators of this person being excellent?
Preferably we're talking graduate or post-graduate level.

We're looking to fill roles of data miners, statistical modeling and data visualization. Thanks Chris, for the suggestion to clarify.
 A: I agree with Chris on most of what he says. Additionally, I'd like to add that without knowing the institutions or universities in detail, just looking at grades would be very misleading. I could easily give a relevant example; I have recently graduated with a masters in engineering mathematics; and taken a variety of statistics courses (with good grades) but I couldnt work in any statistics intensive job right now. That doesn't mean that my uni sucks, but mostly that I didn't manage to learn much out of my statistics courses during university...
Apart from the candidate's knowledge on statistics, I'd also highly value good communication skills; as any cross-disciplinary project eventually boils down to communication problems between experts of different fields. Any test on how well the candidate can share his expertise with others should be a good measure on that.
Furthermore, good computer/programming skills (and no just R is not enough, IMHO) is surely a big plus. If the person has some background in mathematical modeling, it'd be a cherry on the cake :) 
A: It really depends what your company is doing. Are you looking for machine learning experts? Data visualisation experts? Data mining experts?
When I interview statistics PhDs I like to ask them questions about linear regression, as I feel that anyone claiming to be an expert in statistics should at the very minimum be able to explain linear regression to me, and it's surprising how many can't.
Apart from that I'd consider it to be a good sign if they can have a good discussion about model selection/validation procedures, the concept of training and validation sets, cross-validation etc. If they know about classification algorithms (k-NN, SVM, decision trees etc) and can discuss their strengths/weaknesses that's even better.
I find that the particular courses they've studied are rarely a good indicator, and are only really useful for steering the discussion in the interview. If they're claiming to have studied something on their CV, I expect them to be able to discuss it at length.
A: Chris really nailed the data minining stuff.  If you need someone who can also look at experimental data, you can stop all but the most versatile of statisticians dead in their tracks by asking them to explain a split-plot experiment.
