Problem description

CPA firms in one of 3 industry groupings: big4, national, and regional. Each row has a dichotomous measure of IT complexity: low and high as well as an ordinal response 1-7.

Firms may exist in the data set multiple times indicating aspects of the firms, but no indication is provided in the data set about which lines correspond to the same company.

Question of interest is if there are interactions between IT complexity and firm size in ordinal response.

Can we use a nonparametric method?

Consulting response

We are concerned about the lack of information in the data for which rows correspond to which companies.

With an ordinal response (which is essentially a rank transformation of the underlying “response”), it is unclear what an interaction means.

Possibly use an ordinal regression method.

blog comments powered by Disqus


27 September 2018