MGA or Interaction Term

[Sophie]

Hi,

I know that there are two ways to test a moderation via SmartPLS: either in a multigroup analysis (for categorical variables and/ or if more than one relationship is assumed to be moderated) or with an interaction term as part of the model (for continuous variables)

1) I tried to do a MGA with gender, but unfortunately SmartPLS cannot give me the results for the mean path coefficients, standard deviations, t-values or p-values for the women which is the smaller group (n = 92). I do somehow get the original path coefficients for both groups. Is there any way I can still check for the moderating effect of gender? (Perhaps with the inclusion of an interaction term?!)

2) Is it allowed to generate two data groups split by age (younger and older) and then run a MGA? Or is this never recommendable because age is a continuous variable?

Thanks for your help!
Sophie

[jmbecker]

1) Usually, with 92 observations it should be possible to get bootstrapping results. Do you have a variable in your model that has zero variance (all values the same) in your female group? Could it be a very homogenous response (all respondents have indicated the same value)? I would try to dig deeper to understand why this phenomenon occurs.
Of course you could also try to include gender as a dummy variable (0/1) and add interactions.

2) In most situations it is not advisable to dichotomize. However, there is a lively discussion about this issue and situations where the one and the other might be better. A good read is the following special issue in Journal of Consumer Psychology from 2015: https://www.sciencedirect.com/science/a ... 0815000807