one tailed two-sample t-test

[deleted_user]

Hi,
I used proc glm - means/dunnettu to perform a one tailed t-test for a difference between two means. It was wonderful, providing me with upper and lower confidence limits and indicating significance but did not provide a p-value.

So I went into the analyst and simply used the hypothesis - 2-sample t-test for means. It provided t statistics including p-values but I got different upper and lower confidence limits, they don't seem to make sense. Often the lower is a greater value than the difference between means.

How am I misinterpreting this output? Do you recommend a different course of action?

Thank you,
Tyler

[1162]

I often use the LSMEANS statement. You have to use ODS to output both p-values and CLs. Would this work?

proc glm
. . .
lsmeans Group / pdiff=controlu cl;
ods output lsmeans=lsmeans LSMeanDiffCL=LSMeanDiffCL;