Composite null hypothesis: some aspects of F are not determined even when H0 is true. (normal with mean 1) A test statistic T: measures the discrepancy b/w the data and the null hypothesis. We can make large values of T are evidence against H0 in most cases. Explain why?