To add to Chris's question,
If we need to do a transformation in a mediated model, do we have to do anything to the unstandardized beta's before we use them for the Sobel test or can we use them as is?
Thanks,
Andrew
Subscribe to:
Post Comments (Atom)
1 comment:
You should use them as is in the Sobel test and in the comparison of prediction of the outcome from the predictor.
That's why it's critical, if you need to transform a variable, that it be used in its transformed state throughout all analyses. Unstandardized betas computed from transformed variables will be very different than unstandardized betas computed from raw variables. You can meaningfully compare unstandardized betas when both sets were computed from transformed variables and you can meaningfully compare unstandardized betas when both sets were computed from raw variables, but you cannot compare unstandardized betas when one set was computed with transformed variables and the other with raw variables.
Thus, to be a bit redundant (but hopefully clear), if you find that you need to transform your outcome variable in regression 1, you will need to use that same transformed variable in regression 3. Similarly, if you transform your mediator in regression 2, you need use the same transformed variable in regression 3.
Post a Comment