Moderated Mediation when independent variable and moderator are both multicategorical

[Tina Opie]

Dear Stata users,

I am new to Stata (Version 13) having recently switched from SPSS. How can I use Stata to model moderated mediation (first stage model: X --> Mediator is moderated) if BOTH X and the moderator are multicategorical (each variable has three levels)? Another potential challenge is that while X is ordinal (small, medium, large), the moderator is not (poor, excellent, no information present). The no information present is my control condition for the moderator. Below is my data using dataex:

------------------ copy up to and including the previous line ------------------

Listed 100 out of 274 observations
Use the count() option to list more

. dataex Size_Rev Health ATTR EvalProf in 1/20

----------------------- copy starting from the next line -----------------------

* Example generated by -dataex-. To install: ssc install dataex
clear
input double(Size_Rev Health ATTR EvalProf)
1 2 1   6
1 1 0   5
1 1 .   5
1 2 .   .
1 3 .   .
. . .   .
. . .   .
. . .   .
. . .   .
. . .   .
. . .   .
. . .   .
1 1 2   7
1 1 2 5.5
1 1 2 2.5
1 2 3   6
2 2 2 7.5
3 1 1 5.5
3 1 1 6.5
3 1 1 4.5
end
label values Size_Rev Size_Rev
label def Size_Rev 1 "Slim", modify
label def Size_Rev 2 "Average Sized", modify
label def Size_Rev 3 "Overweight", modify
label values Health Health
label def Health 1 "poor", modify
label def Health 2 "excellent", modify
label def Health 3 "no info", modify

------------------ copy up to and including the previous line ------------------

Listed 20 out of 274 observations

I appreciate any and all help.

Thank you!

Tina

[Clyde Schechter]

If you are referring to the approach advocated by Hayes and implemented in the SPSS program Process, there is, to my knowledge, no pre-packaged equivalent written for Stata. However, http://www.ats.ucla.edu/stat/stata/faq/modmed.htm shows how to implement the basic models in Stata code.

Unfortunately, Stata's -sem- command, which figures prominently in these analyses, does not support factor variable notation, so you will have to create your own interaction terms, which is a little tedious to do with multi-level categorical variables. To relieve that tedium and assure that the appropriate combinations are all accounted for, the otherwise largely obsolete -xi- command is helpful. -help xi-.

[Tina Opie]

Thank you Clyde! Yes, I started my analyses in SPSS PROCESS macros; however, I quickly grew weary of creating dummy variables, interaction terms, etc. I'm still hunting for Stata commands that will do group comparisons (for different levels of the multicategorical variables). Thanks again.

[Bruce Weaver]

Re Clyde's comment about -sem- not supporting factor variables (see #2), a quick search of the PDF documentation shows that -gsem- does support factor variables. So if you replace -sem- with -gsem- and fiddle around with the factor variables, maybe you can get something useful. HTH.

[Tina Opie]

Hello Bruce, thank you so much! I really appreciate it! I'm new to the forum but you all are such gracious people with your time. Thank you!