As a follow-up to the last post, I regressed agreeing that one's ethnicity is important (ETHIMP--GSS data) on a list of variables that seemed predictive when looking at the bivariate associations.
The table displays the results. I included p-values since so many of them are just shy of the .05 cutoff. You can read p-values as the chance that the relationship between the predictor and ETHIMP is actually zero.
Believing there should be a law against marriage between blacks and whites is the only independent variable significantly related to thinking one's ethnicity is important. It also has the largest beta. Older people, those living in the South, political conservatives and people who are religious fundamentalists are more ethnocentric (but not significantly so). Gender and church attendance, on the other hand, have high p-values so there is a good chance they are unrelated to the dependent variable.
The results here are stronger here than in the last post probably because the measure of ethnocentricism is more normally distributed. Although the model explains little variance in ethnocentrism (R-squared is only .056) it does suggest that if Americans are encouraged to adopt multiculturalism for themselves, it will take the form of what you would imagine emerging from the South: one that is racialist and conservative.
...and people who are religious fundamentalists are more ethnocentric (but not significantly so)....
ReplyDeleteIf in the quote above you are using 'significant' in the statistical sense, then it is nonsensical to say something is more something, but not significantly so.
Sorry, a pet peeve of mine.
If it ain't significant it ain't significant. That's what the word means.