Using the General Social Survey interface

My post, Who prays more, Democrats or Republicans?, has gotten a little link love. First at Daily Dish, and later at Kevin Drum. That’s great; in fact, a fair number of my GSS based charts get picked up around the blogosphere. But alas, the practice of looking to the GSS to test some intuition or CW hasn’t spread like wildfire. The prayer post took me all of 10 minutes. I’ve posted a link to Berkeley’s GSS interface before, http://sda.berkeley.edu/cgi-bin/hsda?harcsda+gss06, but that doesn’t seem to prod people much, so I thought perhaps this is some behavioral economic conundrum where pictures might induce more interest & initiative.

This is the the circa 1997 style page which greets you after you click the link above.  In “Row” I’ve entered the variable “PRAY,” and in “Column” I’ve entered PARTYID. I’ve put “RACE(1)” into “Selection Filter(s).”  This is all pretty straightforward, I clicked “Run the Table” and produced the data which resulted in the chart where I showed prayer frequency by party affiliation & strength for whites only (because the charts are really ugly, I usually transfer the data into my own spreadsheets).   

How do you find variables?  To the left you see a heirarchical tree which you can explore, or, above you can use the search box to hone in on specific ones based on a query.  Note that the sample sizes can vary a lot for some of these variables, and you can’t always compare them (because they were asked in different surveys).

But that’s not all. At the top left there is a link, “Analysis.” Hover over, and you see this:

As you can see there are lots of analyses you can perform, correlations, comparisons of means, and regressions.  For example, Left-Right political orientation can be converted into a rank order easily.  What’s the correlation between socioeconomic status and political orientation?  You can check:

This isn’t social science. I often do “spot checks” on statistical significances, but for the purposes of posting data on blogs the main criterion for me is that there some value to add beyond the standard anecdotes and “because I said so” arguments (objections to using GSS data do not seem to involve subsequent illustration by critics of better techniques, rather, people simply prefer their own qualitative assertions, intuitions and wisdom, to “faulty data”). Sometimes you will confirm intuition and CW. Sometimes you will have a better quantitative grasp of qualitative intuition and CW. And finally, on many occasions you will see that your intuition can mislead.

Right now only a few weblogs (that I know of!) seem to make regular recourse to the GSS when confronted with a question amenable to inquiry.  The Inductivist, The Audacious Epigone and to a lesser extent Half Sigma. I wish there were more weblogs out there where individuals would take 10 minutes out of their day to double-check their intuitions (on occasion my co-bloggers at GNXP also do GSS data analysis). So with that in mind, if you are inclined to start blogging GSS data, email me at contactgnxp – at – gmail – dot – com. I will add your blog to the blogrolls of both my Gene Expression weblogs, which should give you some Google Pagerank juice. If your posts are topical to the material of Secular Right I would also be inclined to like to you from posts on this weblog.  

I’ve only touched the surface of the various features of the GSS interface. But it really is that simple, and the learning curve is very gentle.