0

I have a dataset with a dependent binary variable and one independent variable (with five categories). When applying logistic regression I have a model that calculates the coefficients of each of these categories in accordance with a reference category (normally it is the first category).

When changing the reference group, we see the relationship between each of the other categories and the new reference group, and so on for the other categories.

I am not that familiar with the thing called contrast but as I understood that it makes pairwise comparisons. However, is it like considered another method ? Is changing the reference group and applying the regression every time is not the same thing ? and in that case what is the right approach ?

kjetil b halvorsen
  • 77,844
An116
  • 367
  • 2
    The difference is that you need to estimate the model each time you change the reference group, which can be inefficient, while contrast avoids that and allows you to test cross-model hypotheses directly. – dimitriy Jan 12 '23 at 22:47
  • 1
    You might be interested in https://stats.stackexchange.com/questions/285210/what-to-do-in-a-multinomial-logistic-regression-when-all-levels-of-dv-are-of-int/544656#544656 – kjetil b halvorsen Jan 12 '23 at 23:40
  • Thank you so much – An116 Jan 13 '23 at 02:21

0 Answers0