sample size in experiments (N of subjects per condition)

[Simone M.]

Hi all,

this is my first post in methodspace. I tried to look for other similar posts, but was unable to find an answer to my question(s).

I am soon going to conduct the experiments for my PhD and have some issues in finding information on sample size for experiments.

 

My questions are:

 

– what’s the rule of thumb for determining the number of people in experiments? How many subjects should I have for each condition (e.g., 30)?

– Do you have any reference for this?

 

(I don’t have any idea on the power of the effects that I plan to obtain so I don’t know how to use to common tools/softwares to determine sample size).

 

Thanks a lot!

Simone

[Rafael Garcia]

Simone, 

Without some estimate of the effect size and the desired power, there is no principled way of determining the needed effect size. 

I would look at related studies in the literature, get an estimate of the effect size there, and use that effect size in a power calculator (like G*power http://www.psycho.uni-duesseldorf.de/aap/projects/gpower/). 

Assuming two independent groups (and the use of a Cohen’s d -like effect size), you’ll select the following options: Test Family – t tests; Statistical test – Means: Difference between two independent means (two groups); Power analysis – A priori: Compute required sample size given alpha, power, and effect size.

Using 30 subjects, you can detect a Cohen’s d of approximately 0.75 (with power of 0.80, alpha of 0.05 and an N2/N1 ratio of 1). If you think the effect will be smaller than that, you’re study will be underpowered.

Hope that helps.

Rafael

[Simone M.]

Dear Rafael,

thanks for your reply. I am not sure I will be able to get a fair estimate of the effect size from previous studies. But anyway I will try to do that.

 

Best,

Simone

[Author]

Posts