Thursday, May 10, 2007

Confessions of an empiricist II

Here I apply my empirical nature to sociology.

I really like sociology for the tools it gives for counting, measuring, watching, and otherwise analyzing human behavior. Over the years, however, I have noticed that I have become increasingly ambivalent about using some of the more sophisticated statistical methods.

Here's why: The more complex the statistical analysis, the more assumptions it makes about the data. These assumptions are not observable, and thus, strictly speaking maybe not empirical.

Ironically, high-end social statistical analysis starts to resemble philosophy--based on what's assumed, not observed. Not that there is anything wrong with either, they are just not my preference.

Of course... this may all just be my reaction to having been introduced to LISREL my first week of graduate school. ;-)


Jerry said...

What do you think of the post-modern critiques of empiricism? As a rather basic version of that: as a composition instructor, I've noticed how my students really tend to value "facts"--meaning quantifiable numbers--over more qualitative measures. Most pointedly, they say that if a source presents these "facts" it can't be biased. Numbers don't lie, they think. But that's not always true, since surveys and studies may slant data through the questions and categories they create. Obviously, it's a big issue, but do you have thoughts on this?

Brad Wright said...

Frankly, I often have no idea what post-moderns are saying... From what I gather, though, I think they go too far in dismissing numbers.

For me the value of numbers is the strength of the sampling design usually associated with them. My beef is more with the thoughtless acceptance of statistical assumptions.

Of course, if you ever want to lose faith in quant data, just watch it being collected. Ugh! So many misinterpretations and various mistakes.