Tukey's HSD Lettered Groupings with Interactions

[cuy01]

Hi All,

I am trying to find a way to get the lettered Tukey's groupings for interaction (much like it would on JMP). I've pasted my code below. Currently with the "LSMEANS ... / adjust=tukey pdiff" statement (where the ... would be the different effects I would put in) I get a chart of p-values comparing all the different levels of the interaction. I've look around online and have been told to use "MEANS ... / tukey" statement. When I tried this, the lettered groupings were given for the main effects but not the interactions. Only the mean values of the interactions were given. I found a macro online for doing this called pdmix800, but it only works for Proc Mixed. Any help on this would be greatly appreciated.

proc glm;

   class Subject Diameter Height Slope Trial;

    model Shld_P1_Abd          Shld_P1_Flex          Wrist_P1 =

               Diameter|Height|Slope subject trial subject*trial

               diameter*subject diameter*trial diameter*subject*trial

               Height*subject Height*trial Height*subject*trial

               Slope*subject Slope*trial Slope*subject*trial/ nouni;

    random subject trial subject*trial

               diameter*subject diameter*trial diameter*subject*trial

               Height*subject Height*trial Height*subject*trial

               Slope*subject Slope*trial Slope*subject*trial/ test;

               LSMEANS Diameter|Height|Slope / ADJUST=TUKEY PDIFF ;

                       run;

[data_null__ · Accepted solution]

What about the LINES option on the LSMEANS statement.

title1 'Randomized Complete Block With Two Factors';

data PainRelief;

   input PainLevel Codeine Acupuncture Relief @@;

   datalines;

   1 1 1 0.0  1 2 1 0.5  1 1 2 0.6  1 2 2 1.2

   2 1 1 0.3  2 2 1 0.6  2 1 2 0.7  2 2 2 1.3

   3 1 1 0.4  3 2 1 0.8  3 1 2 0.8  3 2 2 1.6

   4 1 1 0.4  4 2 1 0.7  4 1 2 0.9  4 2 2 1.5

   5 1 1 0.6  5 2 1 1.0  5 1 2 1.5  5 2 2 1.9

   6 1 1 0.9  6 2 1 1.4  6 1 2 1.6  6 2 2 2.3

   7 1 1 1.0  7 2 1 1.8  7 1 2 1.7  7 2 2 2.1

   8 1 1 1.2  8 2 1 1.7  8 1 2 1.6  8 2 2 2.4

;;;;

   run;

proc print;

   run;

proc glm data=PainRelief;

   class PainLevel Codeine Acupuncture;

   model Relief = PainLevel Codeine|Acupuncture;

   random painlevel;

   lsmeans Codeine|Acupuncture / adjust=tukey lines;

   run;

   quit;

[cuy01]

Thank you! That was exactly what I needed.

[ALSU]

is there any way to obtain graphic tukey separation by means using SAS. I know how to do it in R (PIC BELOW) but I want to learn 

how do to it in SAS since most of my code is in SAS.

[Rick_SAS]

I'm not an expert on this, but see the section of the PROC GLM documentation on "Multiple Comparisons." You can choose from about a dozen different multiple comparison adjustments.  In particular, read the "Recommendations" and "Interpretation of Multiple Comparisons" sections at the end of the doc.

The main graphical presenation of multiple comparisons is the Hsu diffogram, which is created by default when you have ODS GRAPHICS ON and request multiple comparisons. (Or you can request it with PLOTS=DIFFOGRAM on the MEAN or LSMEANS statement.)

For example, examine the output of

ods graphics on;
proc glm data=sashelp.cars;
   class type;
   model mpg_city = type;
   lsmeans type / adjust=tukey;
run;

For a discussion about how to interpret a diffogram and why many researchers prefer it over the graph that you pasted into your question, see High (2011) "Interpreting the Differences Among LSMEANS in Generalized Linear Models"