Hi,
I am working on a project with y = turnover (1/0) which is dichotomous, x= pay (continuous) and mediator = performance rating (could be continuous or categorical).
I am now fitting the model with gsem function in STATA with some confounders and have found out that performance is a mediator.
The only thing I am confused is that in the STATA example, they used
gsem, coeflegend & nlcom
to measure the direct and indirect effect of the model (continuous outcome example). I am wondering if this method will work for EVERY type of gsem fitted?
In another literature Article Direct and Indirect Effects in a Logit Model
the author showed how to use 'ldecomp' to estimate for the effect. However, this method will only work when x is a categorical variable.
I am just wondering if I can use nlcom for estimating the direct and indirect effect of logit model (dichotomous outcome)?
And how should I interpret those coefficients? Are they odds ratios?
Do I have to standardise continuous variables (including confounders) and include them in the glem instead of normal value when doing the mediation analysis?
- Badges
- Science topic
- Similar topics
- Correction