P values for demographic variables in a paper -- which test was performed? I'm reading a paper and they have a table describing the distribution of demographic variables by case and controls, but there is also a Ps column. I assume this is a p-value, but what are they testing for? A significance difference between the distributions? How would you calculate this? thanks, example table below:
Income        control (%) case (%)  Ps
<10,000       38 (23.8)   32 (39.5) 0.0005
10,000-19,999 41 (25.6)   25 (30.9)
>20,000       81 (50.6)   24 (29.6) 

 A: With some judicious searching I have found the paper, it's:
Schiff MA, et al. (2001),
"Serum carotenoids and risk of cervical intraepithelial neoplasia in Southwestern American Indian women",
Cancer Epidemiol Biomarkers Prev., Nov; 10(11): 1219-22.
The p-value you gave in your question is not correct, it should have been 0.005:

(This error made it considerably more difficult to figure out!)
As mentioned in comments above, one possibility for a test there was to test for homogeneity of proportion (/a test of independence) of case vs control across income categories, but I ended up discounting that possibility because the p-value didn't match. But now that we have the correct value from Table 1 in the paper, when we compare with the chi-square p-value calculated on the same counts, we do get a close match:
 chisq.test(matrix(c(38,41,81,32,25,24),nr=3))

        Pearson's Chi-squared test

data:  matrix(c(38, 41, 81, 32, 25, 24), nr = 3)
X-squared = 10.576, df = 2, p-value = 0.005052

As we see, this is the same p-value (to the number of figures given) so it moves from a possibility to a highly likely possibility that this is the test that was performed. [It's possible it was something similar, like a G-test or a Fisher test, I haven't computed p-values for those. The way to be certain that this is what was done is to ask the authors.]
