topic Re: DUNNETS Post-hoc test in Statistical Procedures

DUNNETS Post-hoc test

Achieng — Thu, 03 May 2018 20:08:44 GMT

Hello SAS gurus and my biggest: smiley very-happy:,

I am comparing three outcome groups.

I was using Kruskal Wallis test because we have non-normal data.

 proc npar1way data=ALICEPAPER3MAY2018 WILCOXON;
	 class NewOutcome2015;EXACT WILCOXON / MC;
	  var Age_2005 WeightPA05;
	run;

To do a posthoc test, my supervisor recommended Dunnets.

I have been able to run other tests i.e just palying with SAS and I am successful, i.e Benferroni.

/* Performing the Bonferroni Multiple Comparisons: */

PROC GLM DATA=ALICEPAPER3MAY2018;
CLASS NewOutcome2015;
MODEL Age_2005 = NewOutcome2015;
LSMEANS NewOutcome2015 / CL PDIFF ADJUST=BON;
run;

However, but cant get DUNNETS Test, please help.

Please help and codes as well.

Thank you very much.

Very kind regards

Achieng

Re: DUNNETS Post-hoc test

Cynthia_sas — Thu, 03 May 2018 23:07:26 GMT

Hi:
Did you look in the Doc? There appears to be an ADJUST=DUNNETT option for the LSMEANS statement: http://go.documentation.sas.com/?docsetId=statug&docsetTarget=statug_glm_syntax10.htm&docsetVersion=14.2&locale=en for PROC GLM.

Cynthia

Re: DUNNETS Post-hoc test

Achieng — Fri, 04 May 2018 12:24:41 GMT

Hi, Cynthia Thank you.

I did notice it but was not giving me the correct data I was looking for.
However, after adjusting to.....

lsmeans Treatment / pdiff=controll cl adjust=dunnett;

I was bale to get resulst I want. Thnak you very much.

Can you please direct me to avideo that can help with interpretation.
Thank you.

SAS Output
The GLM Procedure
Least Squares Means
Adjustment for Multiple Comparisons: Dunnett

NewOutcome2015 Age_2005 LSMEAN H0:LSMean=Control
Pr > |t|
MHNWBoth 52.1374046
MHOBoth 49.1176471 0.1778
MetS2015 51.6000000 0.8843

NewOutcome2015 Age_2005 LSMEAN 95% Confidence Limits
MHNWBoth 52.137405 50.521274 53.753535
MHOBoth 49.117647 45.945359 52.289935
MetS2015 51.600000 49.750254 53.449746

Least Squares Means for Effect NewOutcome2015
i j Difference Between
Means Simultaneous 95% Confidence Limits
for LSMean(i)-LSMean(j)
2 1 -3.019758 -7.071581 1.032066
3 1 -0.537405 -3.332869 2.258060