topic Re: How do you code for 1-sided (side=U) in proc glm? in Statistical Procedures

How do you code for 1-sided (side=U) in proc glm?

mgrzyb — Tue, 03 Jul 2018 22:32:32 GMT

HI. I am stuck again. I never had to use a 1-sided test before. I am trying to figure out how to do a one-sided anova or proc glm test.

I tried everything. I'm sorry that is a silly question.

I want to do it on the SIM Variable. There are 18 simulations. (SIM stands for simulations)

My data looks like this: See attachment. Or I have the wide version S1-S20.

Thank you kindly!!

I don't know where that goes. For example, here is GLM code:

proc glm data=elina.widemg; (This is for the wide data of s1-s20)
class gamermg;
model s1-s20 = gamermg/ nouni;
repeated time polynomial / printe summary;
lsmeans gamermg / pdiff cl adjust=tukey;
run; quit;

Re: How do you code for 1-sided (side=U) in proc glm?

PaigeMiller — Wed, 04 Jul 2018 00:32:45 GMT

One sided test of what hypothesis? Please be specific.

The usual test in ANOVA is of the null hypothesis which says the means of all groups are equal. There is really no such thing as a one sided version of this particular tesst.

Re: How do you code for 1-sided (side=U) in proc glm?

mgrzyb — Wed, 04 Jul 2018 04:13:04 GMT

Thank you!!

I did not think there was one. But I read a couple of papers from journals who said that they used 1-sided tests, and I thought they were wrong, and thank you for confirming that.

I review for medical journals, and am particular about what can and can not be done. I think these authors were wrong.

Thank you so very much!!

Happy 4th. I'll be writing all week.

Re: How do you code for 1-sided (side=U) in proc glm?

PaigeMiller — Wed, 04 Jul 2018 11:27:58 GMT

@mgrzyb wrote:

Thank you!!

I did not think there was one. But I read a couple of papers from journals who said that they used 1-sided tests, and I thought they were wrong, and thank you for confirming that.

I review for medical journals, and am particular about what can and can not be done. I think these authors were wrong.

Thank you so very much!!

Again, @mgrzyb, I ask you to state the exact hypothesis of interest. I didn't say it can't be done. I said it can't be done for the primary test, which is the overall F-test, is not and can not be one-sided.

If you are comparing the mean of group 1 to the mean of group 4, of course that one can have a one-sided test.