CONTEXT ADDED BY ADMIN:
END OF CONTEXT
As for the sampling deal, they didn't leave out all the non-white catholics they sampled . . . they just didn't report on those numbers separately because the sample was too small to get reasonable estimates. From my understanding they are included in the total numbers as well as the numbers for Christians . . . they just don't get broken out separately: http://pewresearch.org/pubs/1210/torture-opinion-religious-differences I am beginning to expect these types of polls which cherry pick out subsets of opinions and then portraying them as some sort of "norm" to attempt to try to make some group look bad. It goes back to the old saying of "Lies, damn lies and statistics".
I think that certainly happens with statistics but I don't see it here. I do have problems with the reporting on CNN, and that (in the media) is where most statistical abuse occurs. For example, in this poll, due to sample sizes being relatively small for each group (at most 334, for the regular attenders), the 95% confidence intervals for the quoted stat ("sometimes" or "often" okay, added up) all overlap, indicating no statistical significance at the .05 level. The range for the nonreligious is 34.4% to 49.6%, while the regular attenders is 48.6% to 59.4%. There's a chance, though very slim, that the nonreligious could support torture more, not less, based on the margins of error.
The media often sees things as conclusive when they aren't, and that's how stats are used to mislead. It's unfortunate and most of us don't have the training to see through it as we should.
