topic Re: Summary stat for numberic vars in SAS Programming

Summary stat for numberic vars

Ronein — Tue, 16 Apr 2024 08:49:27 GMT

Hello

I have a list of many numeric variables and I want to calculate for them the following statistics:

Var name

number of observations

number of observations with no missing value

number of distinct value

AVG

SUM

STD

Min

Max

P95

I want that each variable will have a row with the required statistics

I want that the information will be output to want data set

PROC UNIVARIATE data=class outtable=want;
VAR height weight;
RUN;

VARnamenumber_observationsnumber_observations with no missing valuenumber distinct valuesAVGMEANSTDVMAXMINQ1Q2Q3Q4P5P95CV

Re: Summary stat for numberic vars

LinusH — Tue, 16 Apr 2024 09:19:09 GMT

I think that the OUTTABLE= will give you most of the stats OOTB that you are asking for.

SAS Help Center: OUTTABLE= Output Data Set

Re: Summary stat for numberic vars

Ronein — Tue, 16 Apr 2024 09:50:29 GMT

Thanks,

It is perfect,

However I have some questions-

1- I didnt't find calculation of number of distinct values

2- I dont understand how CV is calculated. As far as I know it is STDV/mean but I dont see that this calculation is done

data HAVE;
input A1-A10;
datalines;
.  223  332  138  110  145   23  293  353  458
1   54   61  196  275  171  117   72   81  141
1  170  140  400  371   72   60   20  484  138
1    6  332  493  214   43  125   55  372   30
2  236  222   76  187  126  192  334  109  546
2  260  194  277  176   96  109  184  240  261
2  253  153  300   37  156  282  293  451  299
2  121  254  297  363  132  209  257  429  295
2  152  331   27  442  103   80  393  383   94
1  178  278  159   25  180  253  333   51  225
2  128  182  415  524  112   13  186  145  131
1  236  234  255  211   80  281  135  179   11
2  215  335   66  254  196  190  363  226  379
2  232  219  474   31  139   15   56  429  298
2  218  275  171  457  146  163   18  155  129
1  235   83  239  398   99  226  389  498   18
1  199  324  258  504    2  218  295  422  287
2  161  156  198  214   58  238   19  231  548
2   42  372  420  232  112  157   79  197  166
2   83  238  492  463   68   46  386   45   81
.  267  372  296  501   96   11  288  330   74
.    2   52   81  169   63  194  161  173   54
.  181   92  272  417   94  188  180  367  342
2  248  214  422  133  193  144  318  271  479
2   83  169   30  379    5  296  320  396  597
;
proc univariate data=HAVE outtable=Table noprint;
var A1-A10;
run;
/*_NOBS_='NUMBER OF OBS with no missing value'*/
/*_NMISS_='NUMBER OF OBS with  missing value'*/
/*_SUM_='SUM of values'*/
/*_MEAN_='AVG of values'*/
/*_MEDIAN_='Median(P50) of values'*/
/*_STD_='STD of values'*/
/*_MIN_='MIN of values'*/
/*_MAX_='MAX of values'*/
/*_RANGE_'MAX-MIN'*/
/*_P1_='Percentile 1'*/
/*_P5_='Percentile 5'*/
/*_P10_='Percentile 10'*/
/*_P90_='Percentile 90'*/
/*_P95_='Percentile 95'*/
/*_P99_='Percentile 99'*/
/*_Q1_='Percentile 25'*/
/*_Q3_='Percentile 75'*/
/*_MODE_='Mode value'*/
/*_CV_='STDV/Mean'*/

Re: Summary stat for numberic vars

PaigeMiller — Tue, 16 Apr 2024 10:45:32 GMT

Regarding CV, the formula is right there in the PROC UNIVARIATE documentation.

https://documentation.sas.com/doc/en/pgmsascdc/9.4_3.4/procstat/procstat_univariate_details03.htm

Regarding number of unique values, this is not normally a thing you want for continuous variables, and so this count of unique values is not available from PROC UNIVARIATE.

Re: Summary stat for numberic vars

ballardw — Tue, 16 Apr 2024 13:54:19 GMT

What is your definition of Q4 in this case? Typically that is the MAX value. Q2 is median

I would start with something like this.

proc means data=sashelp.class mean sum std min max q1 median q3  p5 p95 cv STACKODSOUTPUT;
   var _numeric_;
   ods output summary=want;
run;

With Proc Means you also have a number of options related to quantile calculations