effect size negative?

[bubach]

Hello Community,

i need your help. I want to calculate the effect size. I have 4 Constructs (A, B, C, D) with Connection A to B, A to C, A to D and B to D and C to D.

Then i calculated the R-square for the D with all paths, then i removed path A to D and calculate R-square for D again. Now i could calculate the effect size for A.

I added the path from A to D again and removed the path B to D and calculate in the same way the effect size for B. But i was confused, because i got a higher R-square for D without the path B to D. Of course, my effect size was negative.

How can i interpret negative effect size? Or is my way to calculate effect size wrong?

Thanks so much for your help.

Greetings,
Michael

[jjsailors]

But i was confused, because i got a higher R-square for D without the path B to D. Of course, my effect size was negative.

Hi Michael,

In PLS, the computation of constructs is not independent of the paths in the structural model, so as you add/remove structural paths it's not as if the values of the constructs is constant.

That is why you can have the seemingly illogical finding of a negative effect size.

Regards,
John

[bubach]

Hello Mr Sailors,

thanks a lot for your help. So, in this case, R-Square does not depend positively on the number of paths like in multiple regression, right?

Thanks a lot,
Michael

[jjsailors]

R-Square does not depend positively on the number of paths like in multiple regression, right?

That's right Michael, because the values of the constructs are not fixed, they will vary (to some degree) with the form of the structural model.