0
$\begingroup$

I have a couple of conceptual questions about Effect Size, hope someone can guide me here.

I am told that when I am designing my experiment and I want to establish sample size, I need to specify an effect size, and that I can refer to the literature for that.

Now that I reflect deeper:

  1. Given that all experiments are fundamentally different in one way or another (context, etc.), what would be the value-add from transplanting an effect size from a vaguely related paper?

  2. If I'm proposing a programme-based intervention based on certain economics/psychology principles (i.e. and not referencing past similar studies) - how would I determine a viable effect size?

Thanks!

$\endgroup$

1 Answer 1

1
$\begingroup$

A good question, but I am afraid only vague answers can be given. Sample size calculations is a topic full of different opinions.

You are correct, when you say that the difference in experiments can make different studies difficult or completely impossible to compare in a reasonable way. Never the less, together with your subject knowledge that is all you have as guidelines. Often you should be able to say if an effect difference should be about 0.1, 0.5 or 10 based on you knowledge about the intervention. I would recommend to try a few options to get an idea about how much your sample size changes.

I will not neglect the importance and relevance of sample size calculations as guidelines, especially when it comes to reducing patients at risk. However, I would only trust it to indicate if you need 50 subjects, 200 or a 1000 not a detailed number. So if feasible I would recommend to round up gently after the calculation of sample size.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.