Contributor
Posts: 61

# Sample Size Calculation

I want to calculate sample size per group to test proportions if i think and response rate is 1% and i want to detect a difference of 0.5% with power 90%.

I run this:

proc power;
twosamplefreq test=pchi
groupproportions=(0.015, 0.01)
npergroup=.
power=.9;
run;

Is that  correct for what i want to do? I get ~10,400 per group

I tried doing this for response rate 3% and detect a difference of 0.5% with 90% power. Logically i thought the Nper group would be smaller since the response rate is higher but its more. Which makes me think that i am doing something wrong.

Thanks

Super User
Posts: 13,583

## Re: Sample Size Calculation

Did perhaps want the NULLPDIFF option to dest the proportion difference? Set to 0.05 the sample size drops to 86.

Contributor
Posts: 61

## Re: Sample Size Calculation

well i guess then i dont understand what the GROUPPROPORTIONS option means?

Contributor
Posts: 61

## Re: Sample Size Calculation

[ Edited ]

and to be honest with 1% response rate sample size of 86 does not feel reasonable...

Discussion stats
• 3 replies
• 296 views
• 0 likes
• 2 in conversation