In class today we did a 2 proportion z-test on the calculator and by hand. The results were different. At issue was the need for the calculator to have whole numbers for the x1 and x2 values. In this particular problem the sample proportions were 21% and 18% and the sample sizes were 222 and 206 respectively. The students needed to round their n1pl and n2p2 values to enter them into the calculator.
Some students found p-hat pooled, used this to find the standard error and then the standard error to find a z-score. They checked this against the results of using the 2 proportion z-test on the calculator and found a difference.
My question is how the students should present a solution given that they have differing z-scores. I always have them show their work and then let the calculator do the actual multiplication, addition and square root. This question made me realize that there is a problem in doing that. Suggestions?