My psychology grad student and I have a study conducted in three different countries using two different method variants. Neither country nor method variant is a predictor of focal interest, but might be expected to influence our DVs, so we include both variables as factors in our models. In country 1, method A was used, in country 2, method B, but in country 3 both methods were used. I believe this justifies inclusion of both variables, because they are not perfectly confounded, but there are obviously issues here: for example, testing an interaction term would be out of the question.

As it happens, there are no effects of method variant detected for any of our DVs, but country does have some significant effects. My grad student wants to make something out of the country effects in the write-up. I would prefer not to because with a design with only three countries, there is no way to really know what theory accounts for country differences, but that's not really the issue here. My question is - given the partial confound between country and method variant, is it perhaps in fact dodgy to even try and interpret "country effects" on a statistical level, never mind a psychological theory level?

My guess is that the above is enough information to answer the question, but just in case: most of our models have repeated data points per individual, so we nest individual as a random factor in country (all factors are fixed except individual), and method variant is between-subject.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.