In performing and upper-tailed t-test in JMP, the order of the predictor variables determines in which tail the p-value falls. However, the null hypothesis may not directly reflect this outcome. How do you reverse the order of the variables so that JMP does not pick which to give precedence. Example, if I want to perform a t-test for whether the mean of Female test scores are higher than Male test scores (H1: Female-Male >0) but JMP uses the reverse (H1: Male-Female < 0) this alters the order of my null and alternate hypothesis and puts the rejection region in the left tail. Is it possible to control how JMP makes this decision?
Any insights would be appreciated.
One way to specify the hypothesis test as Female-Male is to set the Value Ordering property in the column menu so that the order is Male first, then Female second. In the oneway anova report you will see the hypothesis test under t-test shown as Female-Male.
Agree with Tom -- switching value ordering in the column properties will get you what you want. But that's not strictly necessary -- the t-test you want is probably already there. I ran a t-test on the Big Class sample data set, using Fit Y by X. (Y=height, X=sex).
The two tailed p-value is .0208.
The one-tailed p-value for the null hypothesis that females are taller than males is .0104.
The one-tailed p-value for the null hypothesis that males are taller than females is .9896.
Reject the hypothesis that females are taller than males.
Hope this helps,