Something lost between analysis and reviewer

This is an unfinished tale. About a year ago I did an analysis for human nutrition. I am not going into what the topic of the paper is as it has not been published.  I am going to try and make sense of a reviewers comments.

Lets get this clear, normally reviewers comments are useful, they strengthen a paper and get it stronger. The problem is that I can't relate the reviewers comments to the analysis. He talks of omnibus tests and then comparisons within them. I did no omnibus tests!  I have been through the output several times and they do not exist.

So I am now trying to find what he is talking about. This means sorting through the differences between what I have written in the report, what has been lifted for the paper and what the reviewers comments are. Somewhere in between the three is something I am not getting and I have not been able to place it. Until it is placed I can't deal with what I need to in the analysis.

At present I suspect that the reviewer has seen the raw data and worked out the chi-squared for that, then has gone to the paper and assumed that the chi-squared quoted is from that when in actual fact it is the chi-squared for the adjusted logistic regression.

The alternative is that they have assumed that something is groups when in actual fact they are present and absences. There are two different ways of coding for different traits. If the traits are mutually exclusive you can code them within one variable with each trait being given a different coding category. So for instance if you had favourite colour you might have the options: red, blue, green, yellow, pink, purple, brown and black. You still might get the awkward respondent who said "orange" but that is another category. As you have asked for favourite you do not get the answer "red and blue". However if you as the question "what colour crayons do you have?" then the answer "red and blue" is totally sensible. In other words having a red crayon does not exclude someone from having a blue crayon. So they ate not mutually exclusive categories.

Now when you have mutually exclusive categories you get omnibus tests. When you do not have exclusive categories you do not. Now I can not recall any mutually exclusive categories, all categories could be present or absent in any individual. Indeed I created a category in one part that specifically counted up the number of different categories of other variables that each individual had and split it between those that had only one and those that had multiple. There were about eight different characteristics. That means an omnibus test for every possible category would have 2⁸-1 or 255 different combinations which implies an omnibus test would have 255 different levels.

No thanks.

Oh well maybe my brain will be clearer on Monday and I can find out what is actually going on. That or I am going to ask the prof if the reviewer would mind making their comments clearer.