Invalid numeric data, '<.0001' , at line....

[Rcore]

This message occurs when I use an output data set (SpearmanCorr) in proc corr.

I would like to remove these notes;

15354:NOTE: Invalid numeric data, '<.0001' , at line 101090 column 212.

15355:NOTE: Invalid numeric data, '_' , at line 101090 column 11.

My code looks like this;

ods output SpearmanCorr=mycorrpvalue;

proc corr spearman data=mydata ;

var myvariable;

run;

data new;

set mycorrpvalue;

if myvariable='<.0001' then myvariable=0.0001;

if myvariable='_' then myvariable=.;

if 0<myvariable<0.001 then newvar=0.001;

else newvar=.;

run;

This is because the data set mycorrpvalue contains numbers and characters, '<.0001' and '_'.

How can I totally replace these characters with number?

Sincerely,

Nish

[Astounding · Accepted solution]

The data does not actually contain characters.  In ODS output data sets, p-values typically are numeric, but have a format applied.  (Running a PROC CONTENTS on MYCORPVALUE will reveal that.)  You can remove the format with:

data new;

set mycorpvalue;

format myvariable;

run;

Good luck.

[Rcore]

It worked!  Thank you very much!

[Ksharp]

There are must be some missing value or very small value .

Try round() function to reform your data.

[Kurt_Bremser]

Is myvariable actually character or numeric with a special format?

[Rcore]

Yes, the myvariable had a format.

#

Variable

Type

Len

Format

1

myvariable

Num

8

PVALUE8.4

I could remove it using;

data new;

set mycorpvalue;

format myvariable;

run;

Thank you for your answer!