Power curve of distribution with 2 parameters

#1
Hello.
I want to obtain a power curve for goodness of fit test of a distribution with 2 parameters, let say A & B. My test is actually about parameter B. And say mean=A*B.

So, I'm curious of how should i generated the sample to get the power? Should i generate samples with all the same mean but different B (and hence different A)? OR should i generate samples with the same A but different B (and hence different B)?

Thanks in advanced. And i appreciate any comment regarding this.

-tsukiko-
 
#2
i probably did not understand you question correctly:
-do you want to power curve for goodness of fit test to the data that you are going to generate from the distribution to which you are performing this goodness of fit?