Following problem:

I'm doing a survey where my independent variables are: (1) the perceived attractiveness of a product on a 7-Likert-scale and (2) the willingness to invest in such a product as a numeric value (0-10.000).

During the survey, multiple descriptive questions are asked, such as "level of education", "age", "gender" or "occupation". I'm now doing a multiple ordered logistic regression regarding each of this variables.

Recently I came across the problem of alpha error inflation in multiple comparisons problems. Is this an issue here? I have some weakly significant values, for example Masters students show significantly (p = 0.049) higher results regarding the invested sum - are they really significant or do I have to do a Bonferroni correction (which would, of course, render those results insignificant)?

I'm not sure whether I fully understood the application of the multiple comparisons problem, so I'd be glad if anyone could help :).



1 Answer 1


The fact that it is an ordered logistic model does not alter thins (as opposed to a linear model, a logistic, ...). So if you want to adjust for multiplicity then carry on but if you view these as distinct scientific questions then you perhaps would not adjust.

Note that if you do adjust there are more powerful ways than that named after Bonferroni, Holm's method is worth investigating here.

  • $\begingroup$ Is there a valid method to differentiate b/w distinct and indistinct questions? I feel like e.g. age, occupation, Level of education and so on are somewhat dependent on each other, thus I'd probably account for alpha error Inflation and thus drop those "findings" - does this seem to be legìt? $\endgroup$ Jun 26, 2016 at 22:15
  • $\begingroup$ It depends on the underlying science. For what it is worth I would have said that education and occupation were probably measuring the same thing imperfectly but age is rather less related. But it is really up to you and your prospective audience I think. $\endgroup$
    – mdewey
    Jun 27, 2016 at 13:22

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.