How to get a winsorized mean as a function of time-SAS Univariate

[jacksonan123]

I have the following data with the listed variables for time 1hr.  There are similar data sets for times 2,3,4,6,7,8,12,and 24 hrs.

    ID AMT Time CMT Y
15 1 0 1 3 0.40953 .
16 2 0 1 3 0.2531 .
17 3 0 1 3 30.549 .
18 4 0 1 3 2.2213 .
19 5 0 1 3 1.829 .
20 6 0 1 3 8.8514 .
21 7 0 1 3 29.094 .
22 8 0 1 3 4.2443 .
23 9 0 1 3 0.68437 .
24 10 0 1 3 3.3793 .
25 11 0 1 3 2.8721 .
26 12 0 1 3 0.7927 .
27 13 0 1 3 7.477 .
28 14 0 1 3 0.98951 .

I have done a box plot and ID's 7 and 3 are outliers.  My question is there code that I can use with  SAS univariate that will allow me to obtain the Winsorized mean variable Y for each collection time period?  I  have searched but could not find code modification to do so.  I realize this could be done by brute force by analyzing each time separately.

[PaigeMiller]

So you have separate data sets for each time? Then combine the datasets into one, and perform PROC UNIVARIATE with a BY statement.

[jacksonan123]

This is my code but it says that it can't find Y but I did a proc contents
and Y was the variable name (see below).


Alphabetic List of Variables and Attributes


#

Variable

Type

Len

Format

Label


2

AMT

Num

8

BEST12.

AMT


4

CMT

Num

8

BEST12.

CMT


1

ID

Num

8

BEST12.

ID


6

NAMT

Num

8






3

TIME

Num

8

BEST12.

TIME


5

Y

Num

8

BEST12.

Y

This is my code:

ods exclude all;

proc univariate data=NORM2A winsorized=0.1;

ods output WinsorizedMeans=winMeans;

Var y;

by time;

run;

ods exclude none;

[ballardw]

Copy the code submitted and the messages from the log and paste that into a code box opened using the forum {I} icon.

You do know that exclude all means no output is generated with the example code so your Winmeans data set will not be created.

[Reeza · Accepted solution]

1. Combine them into one data set and identify the source data set as the time.

data combined;
set time.time_hour: indsname=source; *assuming you have a naming convention;

dsin=source;
run;

2. Calculate summary statistics

proc univariate data=combined winsorized=5;
class dsin;
var y;
run;

[jacksonan123]

This code solved my problem