Moderation with a dichotomous and a continous predictor
Moderation means that the causal association between two variables is itself influenced by a third variable. It is tested by analysing the interaction of the supposed predictor with the supposed moderator in their effects on the dependent variable. In this example, a dataset/dataframe called dat
contains three variables, a continuous predictor called independentVariable
, a dichotomous supposed moderator called dichotomousVariable
, and a continuous dependent variable called dependentVariable
.
SPSS
Analysing an interaction in SPSS first requires creating a new variable consisting of the product of the two interacting variables (also see the section on transformation). Here this will be called interactionTerm
. Note that there is debate on whether the dichotomous predictor should be coded 0
and 1
or -0.5
and 0.5
: this influences the interpretation of the resulting coefficients.
COMPUTE interactionTerm = dichotomousVariable * independentVariable.
The regression can then be conducted:
REGRESSION
/DEPENDENT dependentVariable
/METHOD ENTER independentVariable
dichotomousVariable
interactionTerm
/STATISTICS COEF CI(95) R ANOVA.
To order a plot:
GRAPH
/SCATTERPLOT=independentVariable WITH dependentVariable BY dichotomousVariable.
R
R creates the interaction term automatically:
regr(dependentVariable ~ independentVariable * dichotomousVariable,
data=dat);
To also order a plot:
regr(dependentVariable ~ independentVariable * dichotomousVariable,
data=dat, plot=TRUE);