topic Re: Randomly assigning a small category to other categories in SAS Programming

Randomly assigning a small category to other categories

ting1 — Fri, 22 Jul 2022 11:57:01 GMT

I have a dataset from a survey. The question about gender is:

What is your gender?

1 Male

2 Female

3 Transgender

99 Prefer not to answer

The frequency (cell-count) for option 3 (transgender ) is less than 10. I would like to assign value of option 3 (transgender) to option 1 or 2 randomly. Any one has suggestions on how to randomly assigning a small category to other categories?

Thanks for your help in advance!

Re: Randomly assigning a small category to other categories

PaigeMiller — Fri, 22 Jul 2022 12:10:19 GMT

data want;
    set have;
    if gender=3 then gender = rand('uniform')<0.5 + 1;
run;

I add that I am skeptical about this being a statistically valid thing to do, you should be concerned about that and entertain (perhaps) more valid ways to handle this, such as just leaving gender=3 out of your analysis. In the end, you (not anyone else) have to vouch for this being an acceptable method.

Re: Randomly assigning a small category to other categories

mkeintz — Fri, 22 Jul 2022 12:07:18 GMT

Why reassign a small but probably atypical sample to a large category, thereby probably artificially increasing the within-group variance in the variables of interest?

Re: Randomly assigning a small category to other categories

ting1 — Fri, 22 Jul 2022 12:26:45 GMT

Thank you very much for the quick answer.

However, it looks like the method assign all the value of category 3 to category 1. Before running the code

QGENDER
QGENDER
Frequency
Percent
Cumulative
Frequency
Cumulative
Percent
1
1000
49.53
1000
49.53
2
985
48.79
1985
98.32
3
34
1.68
2019
100.00

After running the code:

data want;
set have;
if gender=3 then gender = rand('uniform')<0.5 + 1;
run;
QGENDER
QGENDER
Frequency
Percent
Cumulative
Frequency
Cumulative
Percent
1
1034
51.21
1034
51.21
2
985
48.79
2019
100.00

Is there a way to assign some values of category 3 to category 2?

Re: Randomly assigning a small category to other categories

PaigeMiller — Fri, 22 Jul 2022 13:17:08 GMT

My mistake. I left out parentheses. Try it this way

data want;
    set have;
    if gender=3 then gender = (rand('uniform')<0.5) + 1;
run;

Re: Randomly assigning a small category to other categories

ting1 — Fri, 22 Jul 2022 13:23:45 GMT

Thanks!

That works.

Re: Randomly assigning a small category to other categories

Reeza — Fri, 22 Jul 2022 15:18:06 GMT

@ting1 wrote:

I would like to assign value of option 3 (transgender) to option 1 or 2 randomly. Any one has suggestions on how to randomly assigning a small category to other categories?

From a representation and inclusion standpoint, this is not something that should be done and is in fact the opposite. Please do not do this. If you do put a giant warning around your analysis so that people know you did this and your results are not representative. Also, do not collect information if you don't know how you're planning to use it.

Re: Randomly assigning a small category to other categories

ballardw — Fri, 22 Jul 2022 17:56:37 GMT

I agree with @Reeza . IF I were forced to treat your 3 as a different value on that scale then I would be more likely to combine the 99 and 3 codes. Which would be easily done with a format and not lose any information in the data:

Proc format;

value gender_r

1='Male'

2='Female'

3,99='Trans/ Prefer not to Answer' /* or some other text */

;

Then use that format gender_r with your gender variable for any summary statistics, reporting or graphing purpose in SAS procedures.

Re: Randomly assigning a small category to other categories

ting1 — Fri, 22 Jul 2022 18:36:47 GMT

Thanks for your advice.

Re: Randomly assigning a small category to other categories

ting1 — Fri, 22 Jul 2022 18:37:04 GMT

Thanks!