LSMEANS equivalent when using PROC SURVEYFREQ?

[icipeb]

Good day,

I am analyzing what variables could impact one making a decision (yes or no) based on data from a complex survey design.

Using PROC SURVEYLOGISTIC and the LSMEANS statement, I am able to see if the proportion of people making a Yes decision is statistically different between, let's say, age groups.

The fact that LSMEANS assess all possible combinations (say differences between toddlers to teenagers then toddlers to adults, then teenagers to adults, etc.) is very helpful.

Is there a way to do the same using PROC SURVEYFREQ ?

I know, using TABLE and / chisq, I can determine if a variable, say age, has an impact but I cannot tell if there is a significant difference between the various age groups.

Thanks in advance for your help and suggestions!

Erik 

[PaigeMiller]

LSMEANS only makes sense (to me) in modeling procedures, such as SURVEYLOGISTIC and SURVEYREG.

Its not in SURVEYFREQ because this is not a modeling procedure, and LSMEANS doesn't seem to make any sense here.

What you are asking about, comparing all combinations to one another, in theory can be done in frequency procedures (actually comparing percentages in each cell to other cells), but I don't think SAS does it directly. Here is a paper which contains a macro that does this testing in SAS: https://support.sas.com/resources/papers/proceedings/proceedings/sugi31/204-31.pdf

[icipeb]

Fair enough.

How would you go about comparing a series of means/proportions in the scenario I presented?

Erik

[PaigeMiller]

---

@icipeb wrote:

Fair enough.

How would you go about comparing a series of means/proportions in the scenario I presented?

Erik

---

Modified my original post.

[icipeb]

Thank you, I think I did read about this " compprop" macro elsewhere on the internet.