topic PROC SURVEYMEANS | PERCENTILE Option in SAS Programming

PROC SURVEYMEANS | PERCENTILE Option

_maldini_ — Wed, 02 Nov 2022 23:27:02 GMT

I'd like to get the median, and IQR, of a continuous variable using PROC SURVEYMEANS. I can't seem to figure out the syntax. Neither "50" nor "median" seems to work.

PROC SURVEYMEANS DATA=&dataset PLOTS=histogram PERCENTILE=50;
 	STRATUM sdmvstra;
 	CLUSTER sdmvpsu;
 	WEIGHT &weight;
 	
	VAR 
/* 	DEMOGRAPHICS; */
	age_yrs
	;
RUN;

Is there a

Also, is there a way to add a normality test, like Shapiro-Wilk?

Is there a PROC SURVEYUNIVARIATE? Doesn't seem like it...

Thanks.

Re: PROC SURVEYMEANS | PERCENTILE Option

PharmlyDoc — Wed, 02 Nov 2022 23:42:53 GMT

Use "percentile=(50)"

Or just type "quartiles" in the proc surveymeans statement.

For example:

/*  https://www.eia.gov/consumption/commercial/data/2003/index.php?view=microdata
File 15 */

filename csvFile url "https://www.eia.gov/consumption/commercial/data/2003/csv/FILE15.csv" termstr=crlf; 
 
proc import datafile=csvFile  
dbms=csv  
out=CBECS_2003; 
getnames=yes; 
guessingrows=3000; 
run; 

proc surveymeans data=CBECS_2003 quartiles /* percentile=(50) */ mean;
  strata STRATUM8;
  cluster PAIR8;
  var ELEXP8;
weight ADJWT8;
run;

Re: PROC SURVEYMEANS | PERCENTILE Option

_maldini_ — Wed, 02 Nov 2022 23:47:31 GMT

Sorry, I should have been more clear...I want the median and IQR in addition to the mean. I basically want something that looks like PROC Univariate output so I can assess for normality of the distribution of the continuous variables.

Re: PROC SURVEYMEANS | PERCENTILE Option

PharmlyDoc — Thu, 03 Nov 2022 00:17:14 GMT

The proc surveymeans statement has a range option, but not IQR – SAS must have a reason why.

I suppose you could manually subtract Q1 from Q3.

proc surveymeans data=CBECS_2003 quartiles mean ;
  strata STRATUM8;
  cluster PAIR8;
  var ELEXP8;
weight ADJWT8;
ods output 
Quantiles=Quantiles;
run;

proc transpose data=quantiles name=Quantile prefix=Quantile out=Quantiles2;
var Estimate;
run;

proc sql;
select
Quantile3-Quantile1 as IQR
from Quantiles2;
quit;