Hello everyone,
I have met a problem when testing my hypothesized moderating effects in PLS. All my constructs are measured by reflective indicators, and the reliability and discriminant tests of these basic constructs were fine.
However for the interaction terms (generated by the product-indicator approach as a inbuilt function of this smartpls), the info in the "quality overview" shows that the realibility for some of the interaction terms are poor, with AVE and composite realibility far from acceptable value.
My question is should this be a concern when testing the modetating effects? That is, is the result of the modetating effect testing still valid if the quality of the interaction term is poor?
If not, how can I deal with it? Or should I, instead of using the built-in function of generating the interaction term, create the interaction iterm by hand through calculating the product of the standardized latent variable acore of the predictor and the hypothesized moderator in other sofware first?
Will appreciate very much if anyone can answer...
Best regards,
Hongling