topic Re: PROC TABULATE in ODS and Base Reporting

PROC TABULATE

robertrao — Fri, 19 May 2017 19:01:14 GMT

I want the output like shown below:What changes do i need to make to the PROC TABULATE code:

Proc tabulate data=have;

class source;

var A B C D E;

table (A B C D E ) *source,

sum=

mean=

;

run;

OUTPUT:

Source1 Source2

sum N mean Sum N Mean

Overall --- --- ----- --- --- -----

Re: PROC TABULATE

Shmuel — Fri, 19 May 2017 20:09:21 GMT

Please supply some test data in a format of dataset with datalines (or cards).

Re: PROC TABULATE

Reeza — Fri, 19 May 2017 20:28:45 GMT

Test it until you get what you want. You can't break it...,

Try

Table (A b c d e all), source*(n mean sum std);

As indicated you should provide either test data or use the relevant SASHELP dataset.

In this case, the SHOES dataset is probably closest to your actual structure. But that's really a guess at this point in time.

Re: PROC TABULATE

ballardw — Fri, 19 May 2017 20:32:17 GMT

If you want "overall" to sum the A B C values tabulate does not work that way.

You would have to do the sum in the data set as a separate value. And the "mean" definition may get a bit tricky especially if you are continuing your previous question about getting counts and percents of your "unwanted" values.

Re: PROC TABULATE

robertrao — Fri, 19 May 2017 20:36:09 GMT

data temp;

input meas$ source$ surgery;

datalines;

MEAS_A hosp1 1

MEAS_A hosp2 0

MEAS_A hosp2 1

MEAS_B hosp1 1

MEAS_B hosp1 0

MEAS_B hosp2 1

MEAS_B hosp2 0

;

run;

I want the output like shown below:What changes do i need to make to the PROC TABULATE code:

Proc tabulate data=have;

class source;

var A B ;

table (A B ) *source,

sum=

mean=

;

run;

OUTPUT:

hosp1 hosp2

sum N mean(%) Sum N Mean(%)

MEAS A 8 8 8/8*100=100% 18 19 18/19*100=94.74%

MEASB 1 3 1/3*100=33.3% 2 28 2/28*100=7.14%

Overall 9 11 9/11*100= 81.81% 20 47 20/47*100=42.55%

Re: PROC TABULATE

Reeza — Fri, 19 May 2017 20:39:52 GMT

Why didn't you start with this? And why are there still A B in proc tabulate? And what's actually your Var? The code doesn't line up with your data.

And typing this took the same amount of time as typing an answer would have taken.

Re: PROC TABULATE

robertrao — Fri, 19 May 2017 20:54:15 GMT

Hello,

I am sorry for the confusion. A and B should have been MEAS A and MEAS B respectively.(forgot to change the old code which was suggested by Ballard and Art yesterday).

Shortly after I posted the question I figured out from the article that all the stats have to be in braces to split the results for each of the values under class variable.

http://support.sas.com/resources/papers/proceedings11/173-2011.pdf

Lastly, i could not start with this yesterday since I was asked to split the report after I started working on it..

Hope you will understand..Also I am working on other tools these days so lost the little knowledge I learnt

Thanks

Re: PROC TABULATE

ballardw — Fri, 19 May 2017 20:59:07 GMT

proc tabulate data= temp;
   class meas source;
   var surgery;
   table meas='' all="Overall",
         source='' * surgery='' *( sum n mean="%"*f=percent8.1)
   ;
run;

Note that your original question was showing multiple variable instead of one variable with multiple levels (which is what tabulate works with). the misc ' ' are just to suprees variable names since you didn't provide or indicate a desire for more specific labels other than for mean.

Re: PROC TABULATE

robertrao — Fri, 19 May 2017 21:09:25 GMT

Thanks Ballard...

I need it for multiple variables... That example dataset i provided was a mistake. Can your code be expanded to accomodate for multiple variable ?

Thanks

Re: PROC TABULATE

ballardw — Fri, 19 May 2017 21:45:25 GMT

@robertrao wrote:

Thanks Ballard...

I need it for multiple variables... That example dataset i provided was a mistake. Can your code be expanded to accomodate for multiple variable ?

Thanks

No. Proc tabulate will not sum across other variables.

However you can create a class variable whose values are the names of the other variables and use the example code.

Re: PROC TABULATE

robertrao — Fri, 19 May 2017 21:52:32 GMT

I see. How about using Proc Report instead of tabulate. Can be acheived?

Thanks

Re: PROC TABULATE

ballardw — Fri, 19 May 2017 22:19:52 GMT

@robertrao wrote:

I see. How about using Proc Report instead of tabulate. Can be acheived?

Thanks

I think Proc Report would do the calculation if you are willing to have the A B C as columns and the Source as rows.

I don't use Report often and there are a number of headaches with the way cross-column calculations are done I have to be really desparate to go there when I have an easier way.

It would take about 7 lines of code to make the A B C category data set similar to the one you provided earlier with the variables (or the variable labels for A B C which would likely be better) as a categorical variable.

Re: PROC TABULATE

Reeza — Fri, 19 May 2017 22:47:29 GMT

@robertrao wrote:

I see. How about using Proc Report instead of tabulate. Can be acheived?

Thanks

Or make your data look like it has one variable by using PROC TRANSPOSE. Then you can use the same idea.