Output data and plot

[desireatem]

Can you please help me plot in SAS? So they want to plot the estimated weight loss percentage from regression models (not the actual weight loss) by treatment groups. I have been working on this, but my figure looks wrong. Can you please help? The code I have now is:

proc mixed data=want2 method=ml;

   class mrn drug (ref='0');

   model weightlossn = drug2 time drug2*time / s outpred=Pred;

   random intercept drug2/ sub=mrn;

run;

proc sort data=Pred;

   by  time;

run;

proc sgplot data=pred; 

     series y=pred x=time / group=drug2;

     run;

[PaigeMiller]

So, we don't have your data, we can't see the plot, and even if we could, you haven't explained what is wrong with it. Can you please provide that information?

[desireatem]

Thank you, the plot doesn't come out as two lines.

[PaigeMiller]

Can you please show us the output from PROC MIXED and the plot? Please use the "Insert Photos" icon to include your screen captures in your reply.

[desireatem]

[PaigeMiller]

I did ask for two screen captures, you only provided one.

[ballardw · Accepted solution]

Is your variable weightlossn a percentage of some sort? If not then the Predicted values will be in the original units and type of measure.

If you have a start time measurement then you likely need to create your dependent variable as a "percent change since start" for each of the subjects. Maybe.

Something like this shows what I mean:

data have;
   input id measure time;
datalines;
1   150   0
1   149   1
1   147   2
1   145   3
2   120   0
2   123   1
2   120   2
2   118   3
;

data need;
   set have;
   retain startmeasure;
   by id;
   if first.id then do;
       startmeasure=measure;
       measurepercent=0;
   end;
   else measurepercent = (measure-startmeasure)/startmeasure;
run;

The model statement would use the "measurepercent" variable. Maybe.

If your ID, individual subjects, are using combinations of "drug2" then I have no clue what would really be a good idea.

[Reeza]

Your plot is showing your input data against the estimate for each data point. I don't think that's what you actually want, I suspect what you want is actually more of an effect plot, for example what is the estimate at a set of specific values, not all the input data in your data set.

[Ksharp]

It seems you want mixed effect plot . Try EFFECTPLOT statement and @Rick_SAS wrote this blog before .

[Rick_SAS]

See the article, "Visualize a mixed model that has repeated measures or random coefficients,"

which discusses various options.

I will add that your PROC MIXED code looks wrong: You have DRUG on the CLASS statement, but you use DRUG2 in the model.