How to write the mathematic equation for mixed effect logistic regression results?

48 Views Asked by At

We have a study to predict the disease event outcome: We used the mixed effect logistic regression.

Outcome: binary disease yes(1) or no(0)| Fixed effect: complication1(binary, yes or no), size of hospital (small/medium/big, ref=small), gender (ref=male)| Random effect: hospital (different hospitals)|

Results were:

Random intercept: hospital, covariance estimate=0.03, SE=0.02, p=0.06

Fixed intercept, estimate=-1.32; Fixed effect:

effect Estimate (SE)
Complication1 0.20(0.11)
hospital size (M) -0.07(0.06)
hospital size (L) 1.33(1.10)
gender female 0.14(0.07)

Reviewer asked us to write the formula/equation based on this results above.

Anyone can let me know how to write it based on this outcome which included a random intercept?

1

There are 1 best solutions below

0
PBulls On

This should probably go on Cross Validated, where it would be much easier to write an answer using TeX.

A logistic regression models the probability of disease, π, in the log odds scale:

enter image description here

The equation for your model then becomes:

enter image description here

The dummy indicators and matching parameters can be allocated as you like (as long as they match), for example x1 = Complication1 (so B1 = 0.20), x2 = hospital M (B2 = -0.07), x3 = hospital L, x4 = female. Gamma_i is your per-hospital random effect, which is distributed as follows:

enter image description here

The important parameter is sigma_h or the across-hospital standard deviation or variance, which from your output I'm guessing is 0.03 ('hospital covariance estimate').