Spss correlations n.s. but PLS path coefficients positive

[christinemaier]

Hi,

I just started to work with PLS, so this might be a beginners mistake... I tried to test a mediating effect with PLS. I have two constructs that are related to different outcomes. I wanted to test if the relationships between the two constructs (x1, x2) and the outcome variables are mediated by two other constructs (y1, y2). (My model looks like this: x1 -> y1 -> outcomes; x2 -> y2 -> outcomes)

The problem is, that Spss results show that the correlation between one of the constructs (x1) and its mediator (y1) is not significant (so that the mediating effect could not be supported from the beginning on), whereas PLS shows me high path coefficients and total effects (.54) between x1 and y1. Did I make a mistake? How can it be that they correlate not in SPSS but have high path coefficients? Which result should I believe/continue interpreting?

It would be very nice of you to comment on this problem. Thanks a lot,

Christine