topic Re: compute effect size in Statistical Procedures

compute effect size

Aurélie_D — Thu, 26 Jan 2017 16:20:02 GMT

hello,

I want to compute the effect size of a variable.

mean group1 = 10,978

mean group 2 = 9,228

standard deviation 1 = 0,8402

standard deviation 2 = 0,5767

correlation = 0,59105

I calculated the effecct size with g*power, but the result is 2,56. This is a very high value ?

Can I compute this in SAS with this given information, or am I doing something wrong?

greetings,

Aurélie

Re: compute effect size

Reeza — Thu, 26 Jan 2017 12:35:38 GMT

Do you have access to the raw data? What do you mean by EFFECT size? I'm used to see it in context of regression models not just summary statistics.

Re: compute effect size

Reeza — Thu, 26 Jan 2017 12:39:09 GMT

Assuming this paper is talking about the same thing you are, see the macro included.

http://analytics.ncsu.edu/sesug/2012/SD-06.pdf

Re: compute effect size

Aurélie_D — Thu, 26 Jan 2017 16:21:11 GMT

hello,

thank you for the reaction. I have a dataset with 1 group measured 2 times, so I used a paired t-test. But as you see when I calculate this in g*power I get a huge effect size. The effect size that I get when using method 1 is different than when I use method 2.

METHOD 1

I compute first the effect size in g*power in the additional window (effect size = 2,56).

Then I use this value to calculate the power (power = 1).

METHOD 2

I first calculate the power in SAS (power = 0,9999).

Then I use this power value to calculate the effect size in G*power (effect size = 0,68)

Can somebody explain this?

greetings,

Aurélie