I'm running an independent factoral regression analysis with a cateogrical outcome variable (in R). It's a 2x2 design with two potential outcomes - essentially 'yes' or 'no'. I also have a continuous moderator variable called prog_con_1.

When I run the model, there is a significant interaction between the moderator variable (prog_con_1) and one of the IVs. This IV is called 'conservative_source'.

The odds ratio for this interaction is 2.17.

When I graph the model, I get this...

enter image description here

When I then fit seperate models for each level of the conservative_source IV, I get an odds ratio of 1.2 for when conservative_source is included (the dotted orange line in the graph).

I understand this to mean that for each 1 unit increase in prog_con_1 (the x-axis), the odds ratio should change by 1.2 (multiply by 1.2). What's confused me is that the dotted orange line slopes downwards... doesn't this imply an odds ratio of less than 1?

  • 1
    $\begingroup$ Can you report your whole model output? $\endgroup$
    – Bryan Krause
    Feb 9 at 14:10


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.