Contributor
Posts: 68

Re: Percentiles in proc univariate

Dear All,

Here my problem is i need percentiles of weight and age_months for 0-60 months. so here i need results like weight for Age. in graph X axies age_months and Y axies weight.

For this i tried

proc univariate data=child766;

class Age_months sex;

var weight_kg;

output out=perplts pctlpre=_Pwt P_Age p

pctlpts=6 12 24 36 48 60 ;run;

please suggest me how to go with this??/

Regards,

Anil

Occasional Contributor
Posts: 5

Re: Percentiles in proc univariate

Posted in reply to anilgvdbm

can you please provide the dataset which you are reffering to.

Occasional Contributor
Posts: 10

Re: Percentiles in proc univariate

Posted in reply to anilgvdbm

Try this:

proc univariate data=child766;

class Age_months sex;

var weight_kg;

output out=perplts pctlpre= Pwt_

pctlpts=6 12 24 36 48 60 ;

run;

I suspect that you were specifying too many prefixes.

Posts: 2,655

Re: Percentiles in proc univariate

What is up with the weird pctlpts values?  This is going to give the sixth, twelfth, etc. percentile of all of the data within each month.

And if my hunch is correct, the 60 corresponds to 1he 100th percentile or the maximum.  I would ask the original poster what percentiles of the data they are interested in at each month.

Steve Denham

Contributor
Posts: 68

Re: Percentiles in proc univariate

Posted in reply to SteveDenham

Dear SteveDenham,

Here i am looking for child growth rate  from the age and weight so i need percentiles of each year child's weight. because i cont calculate bmi for child so the from the percentiles i can separate the under weight and normal obese.

Thank you ..

regards,

Anil

Super User
Posts: 10,784

Re: Percentiles in proc univariate

Posted in reply to anilgvdbm

pctlpts=1 to 100 by 1 ;

Xia Keshan

Posts: 2,655

Re: Percentiles in proc univariate

Posted in reply to anilgvdbm

Follow 's recommendation, and get all of the percentiles for each sex and month.  Then you can separate more easily.

However, I think you may want to look at the documentation for PROC QUANTREG, if you are looking for predictors of the various quantiles/percentiles.

Steve Denham

Discussion stats
• 6 replies
• 421 views
• 4 likes
• 5 in conversation