> our statistician tells me that unless I remove the 3-way the 2-way interactions are not interpretable
You may claim that there is some extent of ambiguity when interpreting the two-ways interactions without knowing the specifics of the three-way interaction, but it would be really overstretched to demand on removing the three-way interaction in order to interpret two-way interactions.
> Recode one the variables the other way. Make a new variable (Number2) to have reverse level than
> Number, and then do an analysis with Number2. Check if you get the same results for
> Diagnosis x Complexity as in your first model?
I'm totally lost about this suggestion too. What is exactly the difference between Number and Number2?
Gang
Edited 2 time(s). Last edit at 02/26/2020 10:29AM by Gang.
