You are on page 1of 5

Reporting the Results from a Simple Moderation Analysis

Table 1 below is an example of a table for reporting the zero-order correlations. If you have missing
data on any of the variables in the table, and you took the pairwise option when computing the correlations, the
sample sizes will differ across the cells in the table, which is undesirable. To avoid this, you should, when
computing the correlations, include only those variables that will be included in the moderation analysis (Y, X,
Moderators, and any covariate) and use the NMISS option with Proc Corr. If using SPSS, select “Exclude cases
listwise.”

Your multiple regression will include an interaction term. If it is not significant, report that in text,
something like this: The interaction between messages and resources fell short of statistical significance, F(1,
203) = 0.73, p = .39, R2 = .002. Then you should drop the interaction term from the model, rerun the multiple
regression, and report the results of that reduced model. If your interaction term has only one degree of
freedom (which is typical), then R2 is sr2.

Table 2 below illustrates an efficient way to report the results of such a multiple regression. If desired,
additional columns can be added to report the unstandardized partial slope, sr2, and so on. If any of your
predictor variables, including covariates, are categorical and significant, then you should provide for them the
adjusted means (LSMEANS). For example, if Sex (female = 0, male = 1) were significant, report the lsmeans
on Y for each sex. In SAS these are obtained by using the LSMEANS command with PROC GLM Do
remember to identify any categorical variables in the CLASS statement. In SPSS, use GLM, Univariate.
Identify Y as the “Dependent Variable,” any categorical predictors as “Fixed Factor(s),” and any continuous
predictors as “Covariates.” Under Options, ask for the means the categorical variable(s). If you have two or
more categorical variables, SPSS will, by default, include in the model interactions between/among those
categorical. If you wish to have a main effects only model (common when the categorical variables are control
variables). Click “Model,” “Custom,” and build the model you desire.

If your interaction term is significant, you should probe the interaction. Process Hayes makes this easy
to do. See Table 3 below, and the text that follows the table, as an example of how to present the results of
probing an interaction between two continuous predictors.
Table 1
Descriptive Statistics and Intercorrelations

IAM Income Hours


Variable SWLS WAQ SA AEI AL Age Tenure
Work Bracket Worked
SWLS (.89)
WAQ -.33** (.92)
IAM Work .43** -.44** (.85)
SA -.42** .34** -.80** (.92)
AEI -.19** .33** -.74** .26** (.80)
AL .42** -.34** .70** -.54** -.26** (.78)
Age .10 -.07 .17** -.10 -.16* .13* _
Tenure .11* -.03 .17** -.05 -.19** .13* .57** _
Income
.17* .23** -.02 -.02 .02 -.05 .34** .18** _
Bracket
Hours
-.03 .45** -.07 .01 .05 -.12* .05 .06 .47** _
Worked

M 25.56 70.91 64.10 8.61 15.58 24.28 47.44 11.57 3.95 47.41

SD 6.38 17.61 10.21 5.15 5.38 3.03 11.94 9.45 1.57 10.06
Note. N = 358. Entries on the main diagonal are Cronbach’s alpha. LSS = Life Satisfaction Scale; WAQ = Workaholism
Analysis Questionnaire; IAM Work = Individual Authenticity Measure at Work; SA = Self-Alienation subscale; AEI =
Accepting External Influence subscale; AL = Authentic Living subscale
*p < .05, **p < .001.
Table 2

Predicting Subjective Well Being

Predictor  r

SJAS-Hard Driving Competitive -.035 .131*

Rosenberg Self Esteem .561* .596*

Contingent Self Esteem .092* -.161*

Perceived Social Support .172* .426*

Network Diversity -.089 .134*

Number of Persons in Network .107* .221*


*p  .05

Table 3

Support for Animal Rights Predicted from Idealism and Misanthropy

Predictor  p 95% CI

Idealism .067 .39 -.086, .221

Misanthropy* .303 < .001 .149, .456

Idealism x Misanthropy* -.146 .048 -.290, -.001


*p  .05

As shown in Table 3, misanthropy was significantly related to support for animal rights and idealism
significantly moderated that relationship. This interaction is illustrated in Figure 1. The interaction was probed
by testing the conditional effects of misanthropy at three levels of idealism, one standard deviation below the
mean, at the mean, and one standard deviation above the mean. As shown in Table 4, misanthropy was
significantly related to support for animal rights when idealism was one standard deviation below the mean and
when at the mean (p < .001), but not when idealism was one standard deviation above the mean (p = .14). The
Johnson-Neyman technique showed that the relationship between misanthropy and support for animal rights
was significant when idealism was less than .78 standard deviations above the mean but not significant with
higher values of idealism.
Table 4

Conditional Effects of Misanthropy on Support for Animal Rights

Idealism  p 95% CI

One SD below mean .448 < .001 .236, .661

At the mean .303 < .001 .149, .456

One SD above mean .157 .141 -.053, .367


*p  .05

 Karl Wuensch's Statistics Lessons Page


 Continuous Moderator Variables.
 Complete Example (SAS syntax, output, and presentation of results)
 Psychological Capital Moderates the Relationship Between Emotional Labor and Burnout

Karl L. Wuensch, January, 2020.

PS – I am keeping this document in Word format so that others can copy and paste tables from it to use as
templates for their own tables.

You might also like