I'm trying to use `power twoproportions` to run a two-sided test plotting the minimum detectable effect given a set of potential sample sizes and an assumed control group take-up rate (p_1)
Output:
The resultant image has the phrase
Ha : p2 \neq p1: p2 > p1
Implying that a rejection of the null hypothesis would be defined as having a proportion in the control group that is:
a) not equal to the treatment group
b) less than the proportion in the treatment group
The documentation implies that the code above ought to be running a two-sided test but, as I understand, this ought not condition on part b)
For example, if I add the `onesided` option, then I'm just testing for p1 < p2 .
Question: Is the above code actually running just a two sided test whether p1 == p2 and, if so, why is this second statement included in the output? Is there a separate means by which to run just a two-sided test?
Code:
power twoproportions 0.05, n(2500(500)6000) power(0.8) ///
graph(x(N) y(delta) ///
xtitle(Total Number of Units) ///
xlab(#10, angle(45)))
The resultant image has the phrase
Ha : p2 \neq p1: p2 > p1
Implying that a rejection of the null hypothesis would be defined as having a proportion in the control group that is:
a) not equal to the treatment group
b) less than the proportion in the treatment group
The documentation implies that the code above ought to be running a two-sided test but, as I understand, this ought not condition on part b)
For example, if I add the `onesided` option, then I'm just testing for p1 < p2 .
Question: Is the above code actually running just a two sided test whether p1 == p2 and, if so, why is this second statement included in the output? Is there a separate means by which to run just a two-sided test?
Attached Files
