In logistic regression, can I do a post hoc test by varying the level of the categorical independent variable?

Modeling
1

I have data:
IV: Categorical Variable with 5 levels(circumstance 1, 2, 3, 4, 5)
DV: Categorical Variable with 2 levels(0, 1)

I want to know if the difference between the regression coefficients is significant.
But as I know, it’s inappropriate to compare values directly or use Tukey’s post hoc test in this situation.So, as an alternative, I varied the level of the intercept to see if there was a difference between the five variables.

I wonder it is fine or not and there is a better way.

thank you.!

2

Varying the reference group is fine but for simplicity, you may just compare the posterior distribution (Bayesian analyses) or run a linear contrast (frequentist analyses) for each of the ten comparisons.

1 Like