Knowing that results from significance testing may differ from confidence interval when it comes to proportions, is it valid to 1) based the sample size calculation on CI estimations? 2) make inference based on the resulting CI?
For example, with n=13517, a 1-sided 95% confidence interval for expected proportion of 0.08% will extend 0.04% from it (normal approximation). Based on this information, can I plan a trial with a sample size of 13517? Will I be able to reject the null hypothsis (i.e. p>=0.12%) if the resulting 95% 1-sided upper confidence limit is below 0.12% ? Or does this sample size approach implies that if the oberved proportion (not the UCL) is below 0.12% then the null hypothesis can be rejected?
thank you very much in advance for your thought in this matter.