Use Standardised continuous variables when doing GsEM for mediation analysis?

[Claire Lee]

I am doing analysis to test for mediation effect.
y=dichotomous variable
m=continuous variable
x=continuous variable
I have also got some confounders which are either continuous or categorical.
I know that in moderation analysis, standardised variables should be included in the equation instead of normal values.
But does it also apply in mediation analysis with logistic regression?
Do I have to use Standardised continuous variables when doing GSEM for mediation analysis? (i.e., creating standardised variables of age and include Zage in GSEM instead of age variable)

And would it change the interpretation of coefficient in the output table for regression? Would it have any effect on direct and indirect effect estimates?

[Phil Bromiley]

"I know that in moderation analysis, standardised variables should be included in the equation instead of normal values." Many on this list serve would question this claim. While some authors claim standardizing fixes some of the colinearity problems common with moderators, it is not clear that this actually changes the results (rather than just changing reported colinearity). If it does change the results, then I suspect the variables have been normalized in a way that changes the actual model.

Many of us prefer to stay with the original metrics for variables since we think they make things easier to understand. Standardizing will change the interpretation of coefficients since it changes the scale of the variable.