Latent variable values -- standardised vs. raw :-?

[schifo_pls]

Mellow greetings,


after playing around with group comparisons I stumbled across something that puzzles me.

Let's say I want to calculate two PLS models for women's and men's bodyweight, respectively, based on two separate raw data files (not standardised) with the same measures.

When I choose the original metric and take a look at the LV index values, men are heavier than women, which is fine and corresponds to the data. But when I choose the standardised metric in SmartPLS (Mean 0, Var 1), the resulting LV index value is (much) higher for women than for men :-?

I am aware that this is probably not an exclusively PLS-related issue, still I would be grateful if someone could explain how come :-?


Kind regards & thx in advance,
Mike