Interaction versus moderation effect

noetsi

Fortran must die
#1
I know what these are I think. But I really don't know how to tell them apart in practice. The way I think of it is if there are two predictors X and X1 a moderation effect would be X ->X2 -> Y. For interaction the impact of X on Y varies at levels of X1 [maybe that is saying the same thing].

How would you run moderation in statistical software? I know how to run interaction effects.
 

noetsi

Fortran must die
#2
How would you model this. You have variable with five levels (a variable with 5 types of disability which you can model as 4 dummies showing if you have that disability or not). You think that they effect the impact on the dependent variable. So age has an impact on a DV say at one level (one disability type) but not at another level. Or the impact is in one direction at one disability but another direction for another disability.

How do you model that. An interaction between age and each of the 4 dummy variables?
 

hlsmith

Less is more. Stay pure. Stay poor.
#3
Yup just insert disability into model as cat var, also insert age and the product of age and disability. In SAS you can visualize it if you turn on ODS GRaphics, I did this yesterday looking at age groups * time since covid predicting antibody level.

There is a methods paper in a epidemiology journal in last 5 years on moderation vs interaction. The distinction is confusing. I think one of the following was an author it: Stephen Cole, Daniel Westreich or Tyler Vander weele.
 

noetsi

Fortran must die
#6
Thanks hlsmith although I am not sure I understand that quote. :) I will try to get that book and read the article you posted.
 

hlsmith

Less is more. Stay pure. Stay poor.
#8
Yes I have and it (lsmeans) is. The book is really good. It goes into additive verses multiplication interaction. I wasn't recommending you buy it, but I guess you can.
 

noetsi

Fortran must die
#9
I will get it through the state library. I am sure all of at least 100 people have read it :)

I am still not sure what LS mean really is telling you.
 

hlsmith

Less is more. Stay pure. Stay poor.
#11
I have only used it for scoring data. Thou it stores data as some weird SAS file - I remember that irking me.