A follow up question on: PLS vs. Multiple regression

[sprinner]

Hi,

I have a multiple regression model, but I want to take advantage of PLS. So instead of simply using item averages, I used the latent variable scores from PLS as construct values to put in the regression model.

The correlations I get from multiple regression model are the same as the path coefficient in PLS, and I can get signficance and t value from the regression model. I was wondering is my approach is appropriate, and should I use bootstrapping to get significance and t value, instead of using regular multiple regression?


Thanks,

[viswadatta]

Yes you should be using bootstrapping to get t-values. Why go through the two step process at all?

You also ought to be testing the reliability and validity of the constructs.