proc surveyfreq syntax

[nnn123]

Hi,

I'm currently working with some stratified survey data and want to run out big tables of one set of responses versus another. i.e. how did individuals that answered question 1 subsequently answer question 2, question 3, question 4 etc.

So far i have got this code running producing exactly what i want in a 'program' sheet within SAS EG:

proc surveyfreq data=QUESTION_DATA_AND_STRATA nosummary;

tables Q1 * Q2/ row;
strata LaCode;
weight Q1_Weight;
run;

The only limitation to this though is that this only shows question 1 vs question 2 - I am having to copy it each time for each question. How can i put multiple questions in so I get one big table with more manageable results? Either stacked side by side as columns or underneath each other as rows, it doesn't matter to me. I can't get the syntax right (line 2 is the issue), I've been trying things such as these below with no luck:

proc surveyfreq data=QUESTION_DATA_AND_STRATA nosummary;

tables Q1 * Q2/ row; tables Q1 * Q3/ row
strata LaCode;
weight Q1_Weight;
run;

and 

proc surveyfreq data=QUESTION_DATA_AND_STRATA nosummary;

tables Q1 * Q2; Q1 * Q3/ row;
strata LaCode;
weight Q1_Weight;
run;

[FreelanceReinh]

Hi @nnn123,

Try this:

tables Q1*Q2 Q1*Q3 / row;

Or, equivalently, this:

tables Q1*(Q2 Q3) / row;

[nnn123]

Hi FreelanceReinhard,

Thanks for such a quick response! These codes below create a separate table for each question though. I want them all in one big table if at all possible? There are 22 questions in the survey so it would create 21 mini tables otherwise.

Is this possible?

[FreelanceReinh]

---

@nnn123 wrote:

These codes below create a separate table for each question though.

---

A TABLES request like Q1*Q2 creates a kind of cross-tabulation for combinations of Q1 and Q2 values, so each of the "separate tables" is for a combination of two questions.

Theoretically one could imagine a similar cross-tabulation for a combination of 22 questions (like Q1*Q2*Q3*...*Q22), but in many cases this would be very large, considering that even 22 yes/no questions would have 2**22=4,194,304 possible combinations of answers.

Can you outline the layout of your desired "big table"?

[ballardw · Accepted solution]

---

@nnn123 wrote:

Hi FreelanceReinhard,

Thanks for such a quick response! These codes below create a separate table for each question though. I want them all in one big table if at all possible? There are 22 questions in the survey so it would create 21 mini tables otherwise.

Is this possible?

---

It sounds like you need to create an output data set from Surveyfreq and use another report procedure to control appearances.

Adding something like:

ODS OUTPUT crosstabs=mycrosstabs;

would send the output of the crosstab tables requests to a data set you can manipulate or print, though possibly not the prettiest. I find that "pretty" tables require some manipulation of that data set.