I always wondered how much statistics were emphasized in the physical sciences.
In the stats world, psych stats are still a bit on the simplistic side, albeit, not as simplistic as like percentages or just reporting measures of central tendency. But true experiment is still the gold standard that psychologists strive for, since psych is the social science that most attempts to mimic the physical sciences as closely as possible in its methodology. That's not to say everything can be controlled. Alpha levels for rejecting the null hypothesis are still relatively broad (usually .05) as compared to medical research, for example. Obviously, people are complicated and you can't account for everything that they are going to do or think.
Quantitative sociology starts getting a lot more complicated, since quasi-experiments are much more prevalent. Regression is used in psychology but to nowhere near the extent that it's used in sociology. ANOVAs and t-tests are pretty easy to understand compared to the seemingly bazillion types of regression that are out there. Economics can get really crazy with the predictive calculus models that fill journals. I can generally understand them, but I have to read them REALLY slowly to do so.
I find that SPSS is great for doing basic statistical work, but I always found it clunky when you need to start doing things like transforming variables, regression criticism, doing procedures that aren't built into SPSS's statistical analyses, etc. Systat, which is actually where SPSS got its user interface from, always seemed a lot smoother to me.
'Quantitative Sociology' LOL!
Actually, the psychometrics programs in Psychology departments are at the vanguard of what is going on with experimental design (especially 'real world' quasi-experimental design) and also multivariate statistics. Many sociology departments have been discontinued at quite a few universities. Some programs have kind of sputtered back to life by shifting to the more advanced qualitative training involved in 'marketing ethnography' and other tools that attempt to use classical sociological observation techniques in the market place. Pretty cool actually... kind of like 'living among the lowland gorillas' sort of approach to research. (edit: I have recently completed some projects with these types of sociology-based qualitative research firms, linking time-stamped hard purchase data to coded ethnographic data/behaviors based on direct observations, often camera's set up in grocery aisles. Interesting stuff).
Also, those who don't work with large data sets underestimate the data management part of stats programs. A little program like Systat is fine for the graduate student doing a few regression analyses or whatever. For those of us who work in the data mining field, SAS is still the big dog, since it has massive production capabilities (i.e., macro's and an entire programmable matrix algebra language that is very easy to use). When you are dealing with millions of observations and thousands of variables, you get into a different world on this.
SPSS has most of the market share in the Market Research world, since it is a decent data management package, and has a powerful set of statistical routines, and also because you can disable that simplistic 'point and click' user interface and somewhat mimic SAS's programming power (and SPSS is a LOT cheaper).
Also, having done a post doc in a medical school, +1 that clinical trials for medication/treatments have even more rigid requirements for reducing Type I error, since there is SO much risk.
Finally, to FDeck's comment, for the most part, I try to steer those considering statistics as a career (there is a HUGE demand for applied statisticians who also can do advanced data base management) away from programs in math departments. I always tell them 'for the most part, we can assume Fisher got the F tables right'.! Of course, if you want to break new ground as a theorist, that makes sense, but the real world is looking for statisticians with a good sense of experimental design (especially quasi-experimental design for 'real world experiments'), and good handle on advanced data management (i.e. dealing with massive data sets), and a good tool-box of useful techniques for market research, fraud detection, etc. (i.e., advanced regression, cluster analysis/LCA, perceptual mapping, mixed model ANOVA, multi-level modeling, etc., etc.).
Anyway, funny to talk about this on a bass website, but since it is a statistics thread, kind of fun for me. I started my own applied statistical consulting firm almost 20 years ago after leaving the market research world on the client side, (and academic biostatistics before that) and it has been wonderful. I mainly work with marketing departments, and specialize in real world experimental design (e.g., matched store product tests, advertising test markets, etc.), advanced survey data analysis (conjoint designs for new product optimization, segmentation analyses for consumer targeting, customer and employee satisfaction modeling, etc., etc.).
It has been a great ride, and the field is growing like CRAZY if you have the applied statistical skills combined with some sort of business background, like a Marketing degree, for example.
Whew... OK, up early before starting the work day. Most will find this 'blah, blah, blah', but if anyone is considering a career in this area, feel free to PM me if you have questions.
Edit: Econometrics, as you point out, is another totally different thing. I call it the 'make up for limited data with complixity of analysis' area of statistics!
