Post-hoc command mixed-effects model/multi-level modelling (contrast/margins commands?)

[Sinead Pavlovski]

I am struggling with how to go about using contrast command, margins command, or lincom command for post-hoc analyses.

Parameters:
Random effect: single participant
Fixed effects: 2 categorical (a) "partner" : partner condition 1 = social, 0 = non-social; (b) "cong": gaze congruency 1 = congruent, 0 = incongruent; 1 continuous predictor (c) "lsas": social anxiety (min = 3, max = 139, mean = 56.72412)
DV = coordination (continuous) "intent_cor_bct"


Current commands I have run:
contrast {partner -1 1} {cong -1 1}, small effects
lincom lsas + 1.cong#c.lsas/2, small
di r(df)
margins, dydx(lsas)
lincom lsas, small
estat df


I really need help with 2-way and 3-way contrasts. Does anyone have advice for me?

Current model:
mixed intent_cor_bct i.partner i.cong c.lsas i.partner#i.cong i.cong#c.lsas || id: i.partner i.cong c.lsas i.partner#i.cong i.partner#c.lsas i.cong#c.lsas i.partner#i.cong#c.lsas, technique(dfp 15 nr 3) reml dfmethod(kroger)

[Bruce Weaver]

Hello Sinead Pavlovski. Welcome to Statalist. I'm not going to attempt to answer your questions in this post. Rather, I am just re-posting the commands you used within code delimiters (i.e., the # button on the toolbar). This makes it much easier for other members to read your code. (See item 12.3 in the FAQ for more info.)

Current commands:

contrast {partner -1 1} {cong -1 1}, small effects
lincom lsas + 1.cong#c.lsas/2, small
di r(df)
margins, dydx(lsas)
lincom lsas, small
estat df

Current model:

mixed intent_cor_bct i.partner i.cong c.lsas i.partner#i.cong i.cong#c.lsas ///
|| id: i.partner i.cong c.lsas i.partner#i.cong i.partner#c.lsas i.cong#c.lsas ///
i.partner#i.cong#c.lsas, technique(dfp 15 nr 3) reml dfmethod(kroger)

PS- The /// I included on two lines tells Stata that the command continues on the next line. For more info, see Section 16.1.3 in the User's Guide.