effect size (f2)

This forum is closed, and read-only.
Locked
Ella
PLS Junior User
Posts: 7
Joined: Wed Aug 04, 2010 9:42 am
Real name and title:

effect size (f2)

Post by Ella »

Hi,

in my PLS model I have a dependent variable (Vdep) with 3 independent variables (Vind).

I calculated following effect sizes (f2)

(Vind1) --> (Vdep) f2=1,077
(Vind2) --> (Vdep) f2=0,158
(Vind3) --> (Vdep) f2=0

I`m wondering about f2=1,077. Is there a limit for f2?

Thx!
Ella
User avatar
Hengkov
PLS Super-Expert
Posts: 1599
Joined: Sun Apr 24, 2011 10:13 am
Real name and title: Hengky Latan
Location: AMQ, Indonesia
Contact:

Post by Hengkov »

Hi,

Effect size cannot more than 1.0. I think you wrong compute, please try again or used WarpPLS very easy for it. If effect size high, it's mean collinearity problem.

Best Regards,
Hengky
Ella
PLS Junior User
Posts: 7
Joined: Wed Aug 04, 2010 9:42 am
Real name and title:

Post by Ella »

Thank you!
Ella
PLS Junior User
Posts: 7
Joined: Wed Aug 04, 2010 9:42 am
Real name and title:

Post by Ella »

Hi,

I have still the f squared (f 2) problem. Why f squared can`t be greater than 1? In my opinion it`s not includet in the formular:

f 2 = R 2 inc - R 2 exc. / 1 - R 2 inc.

In my case it`s: 0,728 - 0,435 / 1 - 0,728 = 1,077

Hope so you can help me.

Best,
Ella
User avatar
Hengkov
PLS Super-Expert
Posts: 1599
Joined: Sun Apr 24, 2011 10:13 am
Real name and title: Hengky Latan
Location: AMQ, Indonesia
Contact:

Post by Hengkov »

Hi,

f2 = R2 have range 0-1. Sum f2 = R2 (if you have one predictor f2 = R2), thus not possible f2 greater than 1.0.
If you not wrong compute, I think you model have collinearity.
But I suggest check it using WarpPLS for clear.

Best Regards,
Hengky
Locked