topic Re: quartiles for categorical variable in SAS Programming

quartiles for categorical variable

sfo — Wed, 18 Jul 2018 17:27:29 GMT

Is there a way to divide the categorical into 4 quartiles?

For example I have gender with values F and M and I need to divide all F in 4 quartiles and all M in 4 quartiles.

I know for numeric we can use proc rank but what is the equivalent for categorical?

Thanks

Re: quartiles for categorical variable

ballardw — Wed, 18 Jul 2018 17:42:04 GMT

@sfo wrote:

Is there a way to divide the categorical into 4 quartiles?

For example I have gender with values F and M and I need to divide all F in 4 quartiles and all M in 4 quartiles.

I know for numeric we can use proc rank but what is the equivalent for categorical?

Thanks

I think you need to provide some example data and what you want the output to look like given that example data. You are using "quartile" in a non-standard way so a more specific example will be helpful.

Do you mean to assign the data into 4 groups fore each sex that are roughly the same size? Does the assignment have to be random? If so perhaps:

Proc sort data=have;

by sex;

run;

Proc surveyselect data=have out=want groups=4;

strata sex;

run;

Re: quartiles for categorical variable

PaigeMiller — Wed, 18 Jul 2018 19:15:08 GMT

@sfo wrote:

For example I have gender with values F and M and I need to divide all F in 4 quartiles and all M in 4 quartiles.

Using what criteria would you do this dividing into quartiles? Explain please! Give an example.

Re: quartiles for categorical variable

sfo — Wed, 18 Jul 2018 20:00:14 GMT

Thanks for the solution.

Does it work on 9.2?

I am getting an error:

436 Proc surveyselect data=fdaada_popPK out=want groups=4;
------
22
ERROR 22-322: Syntax error, expecting one of the following: ;, (, CERTAIN, CERTSIZE, DATA, JTPROBS, M, MAXSIZE, METHOD,
MINSIZE, N, NMAX, NMIN, NOPRINT, OUT, OUTALL, OUTHITS, OUTSEED, OUTSIZE, OUTSORT, RATE, REPS, SAMPRATE, SAMPSIZE,
SEED, SELECTALL, SORT, SRSALG, STATS.

Re: quartiles for categorical variable

ballardw — Wed, 18 Jul 2018 21:43:13 GMT

@sfo wrote:

Thanks for the solution.

Does it work on 9.2?

I am getting an error:

436 Proc surveyselect data=fdaada_popPK out=want groups=4;
------
22
ERROR 22-322: Syntax error, expecting one of the following: ;, (, CERTAIN, CERTSIZE, DATA, JTPROBS, M, MAXSIZE, METHOD,
MINSIZE, N, NMAX, NMIN, NOPRINT, OUT, OUTALL, OUTHITS, OUTSEED, OUTSIZE, OUTSORT, RATE, REPS, SAMPRATE, SAMPSIZE,
SEED, SELECTALL, SORT, SRSALG, STATS.

No, the Groups=option was added in 9.3 (I think).

If Randomness is not an issue this is another way:

Using a SAS set you should have as a starting point:

proc sort data=sashelp.class out=work.class;
   by sex;
run;

data work.want;
   set work.class;
   by sex;
   group= mod(_n_,4)+1;
run;

Which will evenly distribute the group identifier within the BY variable.

Re: quartiles for categorical variable

PaigeMiller — Wed, 18 Jul 2018 21:46:38 GMT

@ballardw wrote:

If Randomness is not an issue this is another way:

Using a SAS set you should have as a starting point:
proc sort data=sashelp.class out=work.class;
   by sex;
run;

data work.want;
   set work.class;
   by sex;
   group= mod(_n_,4)+1;
run;
Which will evenly distribute the group identifier within the BY variable.

And if randomness is an issue, then you can assign random numbers to each observation and use those to split the data into 4 groups.