These days, papers are published with higher adjusted odds ratios than the crude odds ratio. Although I consider the reality on the action of the confounders (changing the direction, lowering or increasing odds...), I could not find valid evidence that clearly talks about this issue. Because most younger researchers and master students have a belief to have a lower adjusted odds ratio compared to the crude odds ratio for a given variable during the binary logistic regression analysis. How can we check if the changes in the odds ratio are really due to the confounders' effect or by other technical or procedural errors? Thus, I think it will be important for scholars to discuss it and share the evidence so that researchers have a common understanding on it.
Thank you!
Hello Musulew,
It's always possible that one or more of the "control" variables entered in a LR model acts as a suppressor variable, which could account for one explanation that adjusted ORs (exponentiated LR regression coefficents for a given IV) are higher than crude or unadjusted OR values.
Good luck with your work.
Thank you much Sir/Madam David Morse, for your valuable answer to my question.
More detail on suppression here
https://stats.stackexchange.com/questions/73869/suppression-effect-in-regression-definition-and-visual-explanation-depiction
It is quite well understood
Article The Causal Structure of Suppressor Variables
This article may be of interest:
Article Comparing Coefficients of Nested Nonlinear Probability Models
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS