What is the difference between incremental effect parametization and deviation from mean parametization in logistic regression?
I would like to know without extensive maths explanation.
If any good link to read for that topic, please share me.
This is the same.
"incremental (effect) parametrization" == "effect parametrization"
The fitted coefficients estimate the effects (as log odd ratios) from each level of a categorical predictor to the mean (-> "deviation from the mean").
Given you have a categorical predictor with two levels, e.g. "treatment" with the levels "yes" and "no", the effect parametrization gives you log odds ratios for each level that compare the odds for each treatment group to the average odds (over all treatment groups). To get the log odds ratio of "treated" versus "untreated" you will have to sum up the effects (log odds ratios) of both coefficients (-> incrementing the effect size by summing up the individual "deviations to the mean").
- Badges
- Science topic
- Similar topics
- Correction