I am going nuts trying to work this out....

I'm a Psych. grad student trying to check to see what sample size I would need in the following situations:

1. 1 factor with 4 levels and 10 DVs (repeated measures)

2. 1 factor with 3 levels and 2 DVs (repeated measures)

3. 1 factor with 3 levels, 2 DVS and a covariate which is a categorical IV with 4 levels. (repeated measures)

I need to work out urgently how many participants I need. I have used G-Power but I have found 2 ways to calculate hypothesised sample sizes and the results were totally different.

Any advice would be so gratefully received. I'm really confused.

Thank you , clairealexia.

Below: is a quote from a website that indicates how to calculate ANOVAS, but shoudl we do post hoc when working out sampel size in advance? when I calculated a a priori, the sample size I needed was double (which is a huge problem for me).

Power Analysis for One-Way Repeated Measures ANOVA

________________________________________

Univariate Approach

Colleague Caren Jordan was working on a proposal and wanted to know how much power she would have if she were able to obtain 64 subjects. The proposed design was a three group repeated measures ANOVA. I used G*Power to obtain the answer for her. Refer to the online instructions, Other F-Tests, Repeated Measures, Univariate approach. We shall use n = 64, m = 3 (number of levels of repeated factor), numerator df = 2 (m 1), and denominator df = 128 (n times m 1) f2 = .01 (small effect, within-group ratio of effect variance to error variance), and ρ = .79 (the correlation between scores at any one level of the repeated factor and scores and any other level of the repeated factor). Her estimate of ρ was based on the test-retest reliability of the instrument employed.

I have used Cohen’s (1992, A power primer, Psychological Bulletin, 112, 155-159) guidelines, which are .01 = small, .0625 = medium, and .16 = large.

The noncentrality parameter is , but G*Power is set up for us to enter as “Effect size f2 ” the quantity .

Boot up G*Power. Click Tests, Other F-Tests. Enter “Effect size f2 ” = 0.143, Alpha = 0.05, N = 64, Numerator df = 2, and Denominator df = 128. Click calculate. G*Power shows that power = .7677.