Calculate sample size for linear regression

Question

I have a total of 9 independent variables (IVs) and 1 dependent variable (DV). 2 of the IVs are binary and the other IVs are continuous ranging from 0 to 5. The DV is continuous ranging from 0 to 5000.

My hypotheses look like the following:

IV X1 is positively associated with the DV Y
IV X2 predicts DV Y indirectly through IV X3

I have multiple hypotheses of the first version (1.). The mediator model (2.) is the main hypothesis of my study. I want to use a hierarchical regression to test for my hypotheses.

How can I calculate the needed sample size (using G-Power)?

In G-Power: Do I use 'Linear multiple regression: Fixed model, R2 increase' to calculate the sample size? If yes, what effect size should I use to calculate the sample size?

Canned software isn't going to be good for something like this. I would simulate. — gung - Reinstate Monica, Oct 05 '20 at 01:29
How would you simulate? What do I have to do? Sorry for the question, I am a beginner. — Jensxy, Oct 05 '20 at 07:04
If you're a beginner, you should probably work with a statistical consultant. You can get a sense of how it works by reading the duplicate. — gung - Reinstate Monica, Oct 05 '20 at 12:55
Hmm, thank you for answer. But I don't understand how that helps me to determine the needed sample size for my regression :/ — Jensxy, Oct 05 '20 at 13:03
My apologies, @Jensxy. Simulation is often used for power analyses. The basic procedure is to simulate the data generating processes you believe exist at a given N, & determine the proportion of times it yields significance. Then you can search over Ns. There is an example in my answer here: Simulation of logistic regression power analysis - designed experiments. As I say, this is something that's unlikely to be very doable for a beginner; you'd be better off working with a statistical consultant. — gung - Reinstate Monica, Oct 05 '20 at 16:48

Calculate sample size for linear regression

0 Answers0