As I understand it, power can be used before and after: Before, you use an estimate of effect size and significance rate (eg 0.05) to see what sort of sample you need - ie if between-groups design, how many in each group/condition. Using G-power you can also check what sample size with a low effect size. This might mean you reduce the number of conditions if you think you won't get a sufficient sample size for all your proposed conditions.
Once you have conducted your experiment, you use your result to calculate your actual effect size, and then see if you are underpowered - 0.8 being the standard accepted power (to avoid type 2 error). If you are underpowered, then your result is unlikely to be replicated with a similar sample size.
Just my interpretation so look forward to see other posts
all best
Kate
________________________________________
From: Research of postgraduate psychologists. [[log in to unmask]] on behalf of iljana schubert [[log in to unmask]]
Sent: 04 February 2014 11:14
To: [log in to unmask]
Subject: Re: Mediation Sample Size - Bootstrapping
Dear Jeremy and Ciara,
I am dealing with the same issues and chose a medium effect to work out a sample size. I am not checking effect sizes to see whether the results I got are based on the sample size or whether they are down to something really being there.
My question to Jeremy and the rest of the group is is this a legitimate and good way to back myself up and test whether I was right to assume a medium effect?
Thanks for your help!
Best Iljana
________________________________
Date: Mon, 3 Feb 2014 16:38:22 -0800
From: [log in to unmask]
Subject: Re: Mediation Sample Size - Bootstrapping
To: [log in to unmask]
Hi Ciara
First, remember that power analysis is "Guessing, masquerading as mathematics".
You take a guess at what the effects might be, and then you try to the sample size that would give you that effect.
You don't know what the effects will be, because you haven't done the research yet.
In summary, it's OK to assume a medium effect. But if you don't get a significant result, it might be because the effect was small.
Jeremy
On 3 February 2014 11:03, Ciara gill <[log in to unmask]<mailto:[log in to unmask]>> wrote:
Hey everyone,
I was just wondering if anyone could provide advice on determining sample size for conducting a mediation using product of coefficent approach and bootstrapping. I have the Fritz and MacKinnon paper and have consulted the table at the back but am not feeling at all confident. It notes having a value for each of the effect paths and I just wanted to ensure that you calculate this from prior research. As I am connecting slightly altered variables to what has been previously explored I dont have previous literature for effect size - is it appropriate to assume a medium effect?
Apologies if this is not clear, first gander into this type of analysis
Ciara
This email has been scanned for all viruses by the MessageLabs Email
Security System.
This email has been scanned for all viruses by the MessageLabs Email
Security System.
|