Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- SAS Programming
- /
- SAS Procedures
- /
- randomly selection of records

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

02-11-2016 03:31 PM

i have a file of students and each student can appear more than once in it.

student score

100 A

100 B

100 C

101 A

101 B

102 A

103 B

103 A

103 C

.......

.........

i need to randomly select maximum to 2 records per student

Any idea how to achieve this guys please?

Accepted Solutions

Solution

02-11-2016
04:02 PM

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

02-11-2016 03:45 PM

I think this might get you started. Since you may have some students with only one score if we want to use surveyselect which is designed for random sampling we need to describe how many records to pull for each student so the Proc Freq and Data step create a record for each student and if they only have one score then you get one requested, otherwise 2. Note that the output dataset from Surveyselect, Want, will have additional variables that can be used for weighting and tell what the probability of selection was.

```
proc freq data=have noprint;
tables student/list out=temp;
run;
data samplesize;
set temp;
if count > 1 then _nsize_ = 2;
else _nsize_ = 1;
run;
proc surveyselect data=have out=want
sampsize=samplesize ;
strata student;
run;
```

All Replies

Solution

02-11-2016
04:02 PM

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

02-11-2016 03:45 PM

I think this might get you started. Since you may have some students with only one score if we want to use surveyselect which is designed for random sampling we need to describe how many records to pull for each student so the Proc Freq and Data step create a record for each student and if they only have one score then you get one requested, otherwise 2. Note that the output dataset from Surveyselect, Want, will have additional variables that can be used for weighting and tell what the probability of selection was.

```
proc freq data=have noprint;
tables student/list out=temp;
run;
data samplesize;
set temp;
if count > 1 then _nsize_ = 2;
else _nsize_ = 1;
run;
proc surveyselect data=have out=want
sampsize=samplesize ;
strata student;
run;
```

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

Posted in reply to ballardw

02-11-2016 04:03 PM

thanks ballardw

works perfectly

interesting approach

thx again

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

Posted in reply to ballardw

02-12-2016 05:00 AM

The SELECTALL option provides a simplier solution. No counting required.

```
data student;
input (student score)($);
cards;
100 A
100 B
100 C
101 A
101 B
102 A
103 B
103 A
103 C
;;;;
run;
proc print;
run;
proc surveyselect data=student seed=4354 n=2 out=sample2 selectall;
strata student;
run;
proc print;
run;
```

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

Posted in reply to data_null__

02-12-2016 01:53 PM

Good point @data_null__ I wish I could say that I was providing a more general solution where you had different selection rules based on counts within strata. But I just forgot the SELECTALL option as none of my stratified data gets close to the using all of the records in a stratum.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

Posted in reply to data_null__

02-12-2016 02:46 PM

this is even shorter.

Thanks Data_Null, thanks guys