On 21 April 2012 00:20, Catherine Graham <[log in to unmask]> wrote:
>
> Hi,
>
> Just wondered if anyone could help....i am in the process of writing up my research and i have used Fishers exact test. My research is with children treated for a brain tumour and i have used a theory of mind test. I have coded the results in terms of pass or fail and compared them to a healthy comparison sample. My sample sizes are very small (21 in my clinical and 12 in my comparison).
>
> I have a couple of questions...first to you think it is ok to use fishers exact test with these sample sizes? If so does anyone know how i can estimate the power of my stats? I am trying to use GPower however i am unsure what to enter when it asks for prop1 and prop2.
>
Yes, you can use Fisher's exact test.
Contingency tables are fiddly to power in GPower. You put in the
proportion in each of the cells for the null hypothesis.
So, if you expect 20% of people in your control group to say yes, you put
0.375, 0.125, 0.375, 0.125
In the first column.
Then if you think your intervention will shift that to 30%, you put:
0.4, 0.1, 0.35, 0.15
In the second column.
Then you calculate w = 0.115, and with power of 0.8, 1 degree of
freedom, you need 589 individuals in total.
(That's not very well described, but it's hard. :(
In R it's much easier and more intuitive. The command is:
power.prop.test(p1=0.2, p2=0.3, power=0.8)
And it says:
Two-sample comparison of proportions power calculation
n = 293.1513
p1 = 0.2
p2 = 0.3
sig.level = 0.05
power = 0.8
alternative = two.sided
NOTE: n is number in *each* group
Two-sample comparison of proportions power calculation
n = 293.1513
p1 = 0.2
p2 = 0.3
sig.level = 0.05
power = 0.8
alternative = two.sided
NOTE: n is number in *each* group
GPower is flexible, but not easy.
In addition, if you've collected the data and done the study, what's
the use of the power analysis? If you got a significant result, you
had enough power, if you didn't, you didn't.
Jeremy
|