Hi Stata/Mata users,
I have the following issue. I have to adjust t-tests for size so that their true size equals the nominal size, and then compute the relative p-values. I aim to estimate the power of size-adjusted tests.
What I have done so far does not involte any size-adjustment. In fact, I obtained the t-statistics as follows
m: t = b:/se
and, then, computed the relative p-values
m: pvalue = 2*ttail(N-k, abs(t)).
However, I get size-uncorrected power tests when I compute the rejection rates based on those p-values. So, the naive comparison of powers is completely uniformative.
I would like to know how I have to proceed to get the empirical critical values that must be used to set the rejection rule. That is, is there a function like ttail() which computes the probability that the t statistics - computed as specified above - exceed the critical value of the statistics from the empirical distribution?
Thank you for any advice you can give me.
I have the following issue. I have to adjust t-tests for size so that their true size equals the nominal size, and then compute the relative p-values. I aim to estimate the power of size-adjusted tests.
What I have done so far does not involte any size-adjustment. In fact, I obtained the t-statistics as follows
m: t = b:/se
and, then, computed the relative p-values
m: pvalue = 2*ttail(N-k, abs(t)).
However, I get size-uncorrected power tests when I compute the rejection rates based on those p-values. So, the naive comparison of powers is completely uniformative.
I would like to know how I have to proceed to get the empirical critical values that must be used to set the rejection rule. That is, is there a function like ttail() which computes the probability that the t statistics - computed as specified above - exceed the critical value of the statistics from the empirical distribution?
Thank you for any advice you can give me.
