I want to test whether a hypothesis of whether frailty leads to great outcomes of mortality, i used a chi squared to test the association and there is a significant association, now i am using a logistic regression to test for confounding effects. I have some variables which still give a significant odds ratio for my variable of interest (frailty) however that confounder used e.g gender is giving an insignifcant p-value and a odds of 1 - when put in the model with mortality and frailty - does that mean it isnt a confounder - or do i still report the confounding effects and say for example - the adjusted odds ratio when adjusting for the confounder gender is etc. or is it non significant?
I have attached the output
is the frailty association with mortality void because of the 2 confounders are insignificant?
Boris Sobolev
Perfect Boris, thank you so much!! been searching for this answer everywhere, wish you all the best :)
You have properly controlled for confounding by including them in this analysis. Even though the confounders are not significant they still should be included if the analysis without them shows a different p-value for frailty. If the odds ratios for frailty is different for the analysis that does not include the confounders then you know that affect the odds ratio for frailty, despite their individual small statistically insignificant effect. Confounder should be included anytime they affect the coefficients for the variable of interest.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS