G*power -> MANOVA -> confusion with given effect size

Questions about the implementation and application of the PLS-SEM method, that are not related to the usage of the SmartPLS software.
Post Reply
Tom
PLS Junior User
Posts: 1
Joined: Mon Dec 14, 2020 10:15 pm
Real name and title: T. Arnold

G*power -> MANOVA -> confusion with given effect size

Post by Tom »

Hello together,

I'm working with G*power to calculate the sample size for my experiment.
At the moment I'm a little bit confused handling the effect size values. I have 3 IV (3 groups) and 2 DV => so i will use the MANOVA.

SETTINGS:
I start G*power and select:
Test Family: F-Test
Statistical Test: MANOVA (global effects)
Type of analysis: A priori
alpha: 0.05
power: 0.95

PROBLEM:
G*power proposes Pillai's (V) Effect size f²(V) = 0.0625
What's confusing me is the suggested value of 0.0625!

The variables I found in the literatur for estimating the effect size are f² or eta² resp. eta²(part) for ANOVA/MANOVA etc. I know f, f² and eta² can be converted in each other. But the suggested value in G*power (0.0625) is the well known value for eta² (medium effect). Though the usual value for Cohen's f² = 0.15 (medium effect).
see table below

This really confuses me. Is it a mistyping in G*power (I do not think so) but what is my fallacy? If G*power names it f²(V) I ought to input 0.15 for a medium effect. But then i'll get completely different suggestions for the sample size.
Why uses G*power the values of eta² for f²(V)?

I know the following values for the effect size from literature:
Effect size ·········· small ···· medium ···· big
f ······················· 0.10 ········ 0.25 ······ 0.40
eta² / eta²(part) ··· 0.01 ········ 0.06 ······ 0.14
Cohen's f² ··········· 0.02 ········ 0.15 ······ 0.35

Many Thanks for any help

Tom
Post Reply