Power, Sample Size, Interactions ?

#1
Hello Everyone

BACKGROUND

I am working on a psychology project. Initially, I was going to do two separate regressions for each ethnic group but now I am doing one with ethnicity as a variable - per the committee.

There is a theory that states A and B leads to ---------> C, with things happening in the middle. Such A and B leading to C through the development of (lets call it 1). Also two additional variables (we'll call them 2 and 3) moderate the relationship from 1 to C.

The theory states that it is not ethnicity that is the reason for differences in A and B leading to C, but it is the different ethnic groups have different 2 and 3 (the moderating variables).

So, my committee suggested I do a multiple regression with ethnicity as a variable and it should fall out as a predictor.

They also want we to test the interact of ethnicty with each variable. I hypothesized there should be no interactions between the ethnicity and the other predictor varibles (I hope I was correct in this)

CURRENT QUESTIONS:

1. so in the end there will be the original variables (A, B, 1 (which I think is a mediating variable), ethnicity, and two moderating variables (2 and 3) so far 6 IVs PLUS the 5 interaction terms (ethnicity and the other variables). Grand total 11 IVs and 1 DV. So, I found a power calculator on-line. For Alpha level .05, 11 predictors, and medium effect size f-squares .15 I would need a sample size of 122. Is this 122 for each of the two ethnic groups or122 total meaning 61 of each ethnic group?

2. Even though the theory states there are differences in the ethnic groups amounts of 2 and 3, was I correct to assume no interaction based on the theory? I am confused a little because I dont know how multiple regression actually works to calculate. In my brain if the equation compares matched levels of the variable between the groups, ethnicity should fall out. But, if regression does not work that way and does not account that one group has more 2 or 3 than the other then there will be an interaction. I hope this makes sense.

3. I was going to use Cohen's table for regression sample size but it only went up to 8 varibles. And since I have to do the interactions that brings me to 11 -IVs. I have no idea how to calculate power on my own. I found an online caluculator but I dont think I can reference that one. I also found another one that has a free trial and the company in mentioned in my stats text. Any opinions on some acceptable, citable power calculators, or articles that can teach met to calculate it manually so that I can cite in my doc project. Thanks



Please provide any suggestions!!! *Yay! I just found the email notification setting. I had posted a question months ago and just realized it got answered today, now I will get instant notifications---posting this for those who may not know this*
 
Last edited:
#2
Re: Power, Sample Size, Interactions ? - Resolved I think

:D

By the grace of GOD! I stumbled upon GPOWER calculator! That combined with the Cohen article I was able to calculate the sample size and seems like Cohen said the number would be needed for both groups.

The G Power Calculator is free. Yay! You have No idea the headache that caused me. :yup: