extracting basic measures from proc univariate and reordering them

Reply
Contributor
Posts: 61

extracting basic measures from proc univariate and reordering them

dear Community members,

Happy 2014! Extracting descriptive statistics is among the very first steps of many statistical analyses. I am trying to automate the routine of getting basic measures, such as mean, median, standard deviation and IQR from PROC UNIVARIATE. Unfortunately I got stuck when attempting to extracting data from an ods object called 'BasicMeasures.' By default you will get the following output when using ods output BasicMeasures=:

ods_output_basic_measures_of_proc_univariate.jpg

where the first few columns are your BY-group variables, followed by four basic measure columns- LocMeasure (the highlighted one), LocValue, VarName and VarValue. I would like to create one subset containing the BY-groups+ LocMeasure+  LocValue and one subset containing BY-groups+ VarName and VarValue. As shown in the screenshot of the column attributes, the column name I want to extract is correctly identified but I got an error message 'ERROR: variable LocMeasure and LocValue are not found in the work file. Same error occurs for the other two variables VarName and VarValue. I suspect this may be due to the conflict between the 3 data set options KEEP=, WHERE= and RENAME= in my SET statement, because my codes are working when RENAME= is removed. But I have to rename the two columns in the two subsets so I can stack them later. Any idea? Here are my codes

*step 1: sort data*;

proc sort data=OA_MT_wide; by scoreType scoreValue scoreOrder; run;

*step 2: output basic measures*;

proc univariate data= OA_MT_wide normal;

    by scoreType scoreValue scoreOrder ;

    class scoreOrder;                           

    var age;

    ods output TestsForNormality=     _normality_age;

    ods output BasicMeasures=        _basic_age;

run;

*step 3: get mean, median, standard deviation and IQR and group them into two columns- statType and statValue*;

data basic_age; set

    _basic_age(keep = scoreType scoreValue scoreOrder LocMeasure LocValue

                in=a

                where=(LocMeasure in ('Mean' 'Median'))

                rename=(LocMeasure= statType LocValue= statValue))

    _basic_age(keep = scoreType scoreValue scoreOrder VarName VarValue

                in=b

                where=(VarName in ('Std Deviation' 'Interquartile Range'))

                rename=(VarName= statType VarValue= statValue))

                ;

RUN;

My data are in the excel worksheet 'data'. My codes are also attached (normality_short.sas) for your convenience. Many thanks in advance.rder

Attachment
Respected Advisor
Posts: 3,777

Re: extracting basic measures from proc univariate and reordering them

Change your WHERE data set options to use the RENAMEd variable names.  Where=(statType in('Mean' ...

I would look to PROC SUMMARY to create data sets of summary statistics, should be much easier.

Contributor
Posts: 61

Re: extracting basic measures from proc univariate and reordering them

thank you very much. In proc summary, each basic measure can be explicitly specified and this also produce a wide-format output.

Ask a Question
Discussion stats
  • 2 replies
  • 523 views
  • 3 likes
  • 2 in conversation