Study design question- adjust for other exposure variables or examine subsets?


TS Contributor
the first study has some suspicious indications of p value fishing. The only significant result that they found, was for the age group of women over 20,however there was no indication (at least in the abstract) of including age as a factor into the study.And what is so magical about the mothers age of 20 years? And how come age is only important for a combined exposure?

As for the adjustments, I would think that using three indicator variables like H 1 if exposed to herbicides 0 not exposed and testing for all interactoons would have been a better choice .

As for the second one, it was not clear to me from the abstract whether they actually ran many models (one per each chemical) or just one per a class of chemicals ("using broad classes" is a bit ambiguous). Running one model per each chemical would raise several issues - like what would be the sample size for these tests, surely a LOT less then what was reported for the whole study, right? Also, that would increase the chances of false positives by a huge margin - as they did not see any positives I rather doubt tjat tested per chemicals.

I hope this helps a bit