View Single Post
Old
08-26-2012, 12:29 AM
  #63
Patman
Registered User
 
Join Date: Feb 2004
Posts: 328
vCash: 500
Hello all, occasional HF lurker myself but I thought I'd put in my two cents.

My name is Patrick and I hold a Ph.D. in Statistics and work for the federal government. My main work relates to problems in surveys specifically using "small area estimation" but I'm still young so I'm more at the "any port in the storm and whatever gets me a paper". I'm more of a modeler than a sampling-based estimation person.

My main research involve "small area" (hierarchical) models, Bayesian methods and techniques (I'm more of a pragmatist), with some knowledge and interest (waning) in spatial methodologies, and profess an interest in generalized linear models (Poisson, Binomial, etc.) So, in principle, I am more of a generalist... which I've been told isn't useful. Tough being a researcher who is more interested in product dev and getting useful stuff done. But I suppose researchers aren't supposed to be useful.

I don't spend much time doing recreational sport statistics unless somebody poses me a really good problem and I want to work on it. I'd rather be working on my post-Ph.D. life... I worked to have fun... my turn.

I'm usually a good sounding board for things but as a Ph.D. statistician I'm not a fan of canned analysis or doing stuff because you can. Just because you have a chainsaw doesn't mean you should use it to build a birdbath.

The main sources of issues in sports analytics is data access. A lot of gate keepers out there and many of them want $$$. After all, if you can give somebody a competitive edge somebody will want to do something with it.

I've got some toy problems I wouldn't mind working up just as a matter of pushing the question but nothing I would consider hard and sound analysis. For instance, presume a Poisson model and assume its correct (yes, assumptions... they're needed) then when do you pull the goalie? I've programmed this one up before and the question is a bit more open-ended than one would think... but I think I'm not the only one who has come to this conclusion. On the other hand, nobody has tied the bow on it and presented it... I think... I'm behind on my reading of JQAS. Nevertheless, simplistic engineered exercises can sometimes tip the worldview.

My opinion is those who will push the problems along will be a person good at data processing working with somebody who has a measured hand at analytics. Keep things simple and don't promise the world.

i think there's a few good opportunities for high-level bachelors or masters students to work on a toy problem or two. Don't expect the world out of it, but use it as a vehicle to get better at operating with data and programming statistical mathematics (SAS, R, C, Fortran, or otherwise...).

Keep it a healthy hobby. Nobody should try to expect to recreate Shakespeare. If you do, don't try to go through peer review... and don't call it shakespeare. Especially if it isn't. Which reminds me... I have a referee report to submit.

Patman is offline   Reply With Quote