Hello everyone,
I'm using Stata 12.
I'm conducting a moderated regression (2 categorical variables, 1 continuous) for which the 3-way interaction effect is significant. I probed this interaction (i.e. tested simple effects) using the contrast command. My colleague is a social psychologist and asked if my probe for interactions & subsequent slopes and graphs generated from my analyses are based on simple slopes analysis that use values 1SD above and 1SD below the mean. This is typically how it is done in social psychology.
I've never done this 1SD above and below the mean interaction probe before, although I know it was proposed by Aiken & West (1991). So I'm not exactly sure whether what I have done in Stata (testing for simple effects) is the same as simple slopes analyses. Is it? The Stata results aren't based on 1SD above and below the mean though. Stata gives a slope (of continuous variable) for each combination of the categorical variable.
If simple effects and simple slopes analysis are not the same; could someone please help me understand how to conduct the latter? If they are the same-- I'd like a lay man's explanation that I could use to express this to my colleague. I know what simple effects are. My lack of understanding of simple slopes analysis is what's confusing... My syntax that I used to test simple effects in Stata is below. Thank you so much!
regress psat i.sex##i.treat##c.mc /* moderated regression*/
contrast i.sex#i.treat#c.mc /* test the significance of the interaction*/
margins sex#treat /*compute adjusted means by sex and treatment groups for mc)*/
marginsplot, bydimension(treat) noci /*plots fitted values of stereotype perception as a function of sex and treatment group*/
margins, dydx(mc) over (treat sex) /* computing stereotype perception slope for each combination of sex and treatment*/
contrast sex#c.mc@treat, nowald pveffects /*simple effects: comparing males & females in terms of stereotype perception by treatment group*/
Katherine Picho
I'm using Stata 12.
I'm conducting a moderated regression (2 categorical variables, 1 continuous) for which the 3-way interaction effect is significant. I probed this interaction (i.e. tested simple effects) using the contrast command. My colleague is a social psychologist and asked if my probe for interactions & subsequent slopes and graphs generated from my analyses are based on simple slopes analysis that use values 1SD above and 1SD below the mean. This is typically how it is done in social psychology.
I've never done this 1SD above and below the mean interaction probe before, although I know it was proposed by Aiken & West (1991). So I'm not exactly sure whether what I have done in Stata (testing for simple effects) is the same as simple slopes analyses. Is it? The Stata results aren't based on 1SD above and below the mean though. Stata gives a slope (of continuous variable) for each combination of the categorical variable.
If simple effects and simple slopes analysis are not the same; could someone please help me understand how to conduct the latter? If they are the same-- I'd like a lay man's explanation that I could use to express this to my colleague. I know what simple effects are. My lack of understanding of simple slopes analysis is what's confusing... My syntax that I used to test simple effects in Stata is below. Thank you so much!
regress psat i.sex##i.treat##c.mc /* moderated regression*/
contrast i.sex#i.treat#c.mc /* test the significance of the interaction*/
margins sex#treat /*compute adjusted means by sex and treatment groups for mc)*/
marginsplot, bydimension(treat) noci /*plots fitted values of stereotype perception as a function of sex and treatment group*/
margins, dydx(mc) over (treat sex) /* computing stereotype perception slope for each combination of sex and treatment*/
contrast sex#c.mc@treat, nowald pveffects /*simple effects: comparing males & females in terms of stereotype perception by treatment group*/
Katherine Picho
Comment