Interpretation of category-specific effects

Hi,

This is the first time I use sratio() with cs(). Three of my predictors are of type ordered categorical with four levels but when I check with summary() I see five parameters (b_val is one of the ordered categorical variables):

b_val.L[1]       0.30      0.06     0.19     0.41 1.00    19734     7825
b_val.L[2]      -0.05      0.06    -0.17     0.06 1.00    18618     8281
b_val.L[3]       0.22      0.07     0.09     0.35 1.00    19817     7319
b_val.L[4]      -0.12      0.08    -0.27     0.03 1.00    23470     7530
b_val.L[5]      -0.06      0.09    -0.24     0.12 1.00    22062     7429

@paul.buerkner and Vuorre write in their paper Ordinal Regression Models in Psychology: A Tutorial:

… one can model predictors as having category specific effects so that not one but K coefficients are estimated…

but I can’t seem to find an explanation on how to interpret the parameters above. Any input much appreciated since this is the first time I use cs()!

In the sequential model, every transition from one to the next category is modeled as a separate latent variable as explained in the cited paper. Thus for cs() terms, the 1st coefficient expresses the effect on the latent variable deciding between the first and the second category, the 2nd coefficient expresses the effect on the latent variable deciding between the second and third category, and so on.

A very big thank you, @paul.buerkner. I should’ve understood that since my outcome is ordered categorical with 6 levels. I feel extremely stupid, but that’s something good since I learned something new…

No need to feel stupid. It is important that we all keep asking questions. :-)

1 Like