Friday, December 12, 2003

all i'm saying is, it has to do with nuns...

I just did a little hypothesis testing of my own here, of an idea that I had way back in graduate school but never tested. The numbers in the "Mean" column below represent the proportion of correct answers given to each of the ten items on the vocabulary test administered to some of respondents of the General Social Survey (a nationally representative survey that has been collected either annually or biennially since 1972). Separate numbers are presented for Protestants and Catholics.


Variable | Mean N
worda | .8278244 12162
wordb | .94318 12214
wordc | .2581398 9429
wordd | .94522 12249
worde | .7758351 11496
wordf | .8072227 11630
wordg | .3252313 11026
wordh | .3539879 9554
wordi | .7703412 12192
wordj | .224678 11492


Variable | Mean N
worda | .8381347 4825
wordb | .947601 4752
wordc | .3018917 3859
wordd | .9399875 4799
worde | .7709067 4544
wordf | .8415969 4659
wordg | .4609026 4476
wordh | .3661499 3870
wordi | .8036644 4803
wordj | .2301779 4553

Notice that the biggest difference between Protestants and Catholics is for wordg, where Catholics are about 40% more likely to give the correct answer to that item than are Protestants. So, the two questions: (1) why did I make these separate tables for Protestants and Catholics? and (2) what's up with wordg? The words that the GSS uses for the vocabulary test are not available in the GSS codebook or any other public place, out of fear that putting the words in a public place will cause them to spread like wildfire through the general population, as millions are dying to know what these words are so they can be better prepared for a vocabulary test they will almost certainly never take and will have no consequences for their lives even if they do. Anyway, through means I cannot disclose, I actually have the list of words the GSS uses. When I first looked over the list of words, I thought to myself, I bet Catholics have an advantage on this one word, because it's a word I commonly associate with nuns (sorry, can't say what it is).

So, just now, I happened to have the cumulative GSS file up in Stata for different (although not entirely unrelated) purposes, and I decided to see whether there was a relative edge for Catholics on any of the words. I made the table above, and, lo, the Catholic advantage on wordg turned up. Only then did I go to my bookcase and dig up my copy of the vocabulary test to see what wordg was. Sure enough, wordg was the word with nunnish associations.

To my knowledge, no one has ever noticed this before. Real science, real findings, right here live on this weblog!

No comments: