Suppose you are doing an experiment, and have 2 samples, 1 for proportion in control and another for proportion in test. My stats teacher said we need to do a difference of 2-proportion hypothesis test and construct a normal #total successes n1 n + n curve with p-hat being the pooled proportion ( ) and standard deviation (s) being p(1 - p) PT-PC ( 1 1 + n '1 n S And then z-score is calculated using Question: But why should the normal curve used for hypothesis testing be constructed using statistics from both control and test samples; why not just from control and turning this into a 1-proportion testing?