How to compare groups within a group with a t-test?


Hey guys,

I want to find out if females (blue line) significantly are different from the men (red line). For "normaal gewicht" (normal weight) the confidence interval don't overlap, which indicates that they significantly are different (n=878). But for the other two groups (overweight (n= 138) and obesity (n= 29) the confidence intervals overlap. The intervals are bigger, because their n is smaller. My total n = 1045.

My real question is:
How can I do a test in SPSS where I can find out if there are significantly differences between males and females (Moderator) within their weight group (independent variabel) on their sociometric status (dependent variable)


TS Contributor
You can use "select cases" and then perform a t-test for each group. Or you can use "split file" with the
weight group as splitting variable and then perform t-tests.

Or, under GLM you perform a univariate analysis of variance, with the factors weight group and gender.
The interaction between gender and weight group represents the (potential) moderator effect. Although
your description is not one of a moderator, but just of a main effect.

With kind regards



Less is more. Stay pure. Stay poor.
In addition, since you are comparing the sexes across three groups, your risk of false discovery goes up and an alpha correction may be warranted.

Also, if you have their actual weight or BMI values using it as a continuous variable instead of categories may provide some additional information about the sexes and could be preferred.