Why don't more people work as Stata programmers?

[skolenik]

I thought I'd share -- stumbled upon on linkedin. Most of that applies to Stata programming, of course (except that Stata is not object-oriented, but rather is a procedural environment.)

http://www.forbes.com/sites/quora/20...s-programmers/

[Joe Canner]

I agree with most of this, and particularly resonate with the personality trait section. As an undergraduate in computer science in the early 80s it was obvious to me that a lot of people were trying to get into CS for the good job prospects, but many did not have the logical thinking skills and other personality traits required.

I am curious, though, about the "CS programs do not teach programming. CS programs teach theory" statement. When I was in college, we learned both. Granted, the languages we learned were not the ones that we ultimately used in the real world (mainly because the school couldn't keep up with the real world), but learning a language along with the theory gave us the tools to learn whatever the language flavor-of-the-month happened to be. I have a hard time believing that CS programs no longer teach programning, but only theory. Anyone else have more recent experience on this?

[skolenik]

Joe, in reaction to "[ABC] programs to not teach [XYZ]. [ABC] programs teach theory" -- I can safely say that I don't use anything from my statistical theory training now in my job. I learned pretty much every skill that I apply on a daily basis (including, of course, Stata) outside of the major that is written on my fancy golden print diploma. A regression analysis class, with residual diagnostics, model selection and such, may be the only exception. Stochastic calculus or point processes, while surely fancy, did not help me much with neither data cleaning nor non-response analyses. (If I were employed on Wall Street, I might have felt differently about these courses, of course ). So by the same token, statistics programs do not teach data analysis. Statistics programs teach theory.

[Nathan E. Fosse]

Thanks for this, Stas Kolenikov:

My experience particularly resonates with this: "The truth is programmers are artisans and to get the best results from a project the wise thing to do would be to ask the people who are experts at programming how things should be done!" Amen!

Here's my sociological take on the question, "why are there not more Stata programmers?":
(1) demographic/disciplinary: R has a larger pool of programmers because it is more likely to be used by programming folks. Stata is an great hybrid that bridges GUI "lock down" software and the wild programming West of R/Python, etc.
(2) functionality: for me, R demands programming for general use; personally, I know many find it more difficult to use, but I find the object-oriented programming to be very intuitive.
(3) feedback loops: software and cognition cultivate different styles of thought. SPSS folks, I notice, prefer to "see" the data (a la Stata's "browse" command). Again, this is anecdotal, but I think, because Stata's interface is hybridized, it permits different "styles of thought," without scaring users into think one must be a master programmer (this, despite, R's GUI interfaces). Stata is still, in my view, one of the best programs to accommodate a wide range of thinking types: one may use it purely GUI, purely in the command interface, or one may dig deep into the programming functions.

My 2 cents,

- Nate

[Joe Canner]

Stas,

I think I would agree with you regarding statistical training (in contrast to computer science training). I work with a lot of students and former students who have taken a number of graduate level statistics courses but who have very little practical training in statistical programming, data cleaning, etc. Most statistics courses use small, contrived example data sets which require very little preparation before doing the analysis that illustrates the statistical topic at hand. It is indeed in the day-to-day tasks that one obtains these more mundane, albeit essential skills.

Regards,
Joe

[wbuchanan]

Truer words were never spoken. I can definitely empathize quite a bit with the article. I only had two real stats/methods courses and Stata was not a huge part of it, so I ended up teaching myself more about stats and Stata. In terms of the languages, if nothing else the

if

statement in commands would keep me glued to Stata for a long time. R is getting a bit better with incorporating conditional statements in function calls, but the

if/in

statements are truly things of beauty and incredibly intuitive. Although Stata is pretty inexpensive when lined up against the competition, some of the wave of R use is probably a bit of a cost factor, but getting someone up and running in R would definitely take a decent amount longer if they have limited experience with programming languages. I'd be interested to hear James Fielder's take on all of this from the perspective of using Python.