topic Re: Delete negative values ( NOT '.') in SAS Procedures

Delete negative values ( NOT '.')

spraynardz90 — Thu, 12 Jul 2012 07:57:04 GMT

Hi I am looking to delete any observation with a negative value...

Generally I use:

data x; set x; if variable<0 then delete; run;

However this deletes blank observations such as '.' along with any negative values.

Does anyone know how to delete only negative values?

Unfortunately the statement below doesn't work:

data x; set x; if variable=- then delete; run;

Re: Delete negative values ( NOT '.')

shivas — Thu, 12 Jul 2012 08:45:43 GMT

Hi,

Just try this....I am not sure this is optimized solution or not.

data one;

input nub;

cards;

-20

-10

;

run;

data two;

set one;

if index(nub,'-') then delete;

run;

Thanks,

Shiva

Re: Delete negative values ( NOT '.')

Ksharp — Thu, 12 Jul 2012 09:20:29 GMT

You need not missing() to judge it .

data one;

input nub;

cards;

-20

-10

;

run;

data two;

set one;

if nub<0 and not missing(nub) then delete;

run;

Ksharp

Re: Delete negative values ( NOT '.')

SteveDenham — Thu, 12 Jul 2012 11:07:52 GMT

How about:

data one;

input nub;

cards;

-20

-10

;

run;

data two;

set one;

if .<nub<0 then delete;

run;

As . is always less than any value.

Steve Denham

Re: Delete negative values ( NOT '.')

data_null__ — Thu, 12 Jul 2012 11:14:22 GMT

SteveDenham wrote:

As . is always less than any value.

Are you sure?

2360 data _null_;

2361 x = . min ._;

2362 put x=;

2363 run;

x=_

Re: Delete negative values ( NOT '.')

SteveDenham — Thu, 12 Jul 2012 11:24:11 GMT

How about ". is always less than any value on the real line (-∞,∞)" ?

And I learned something new about the underscore--I always thought it to be a character only.

Thanks!

Steve Denham

Re: Delete negative values ( NOT '.')

data_null__ — Thu, 12 Jul 2012 11:51:54 GMT

It is a special missing value not character underscore.

Re: Delete negative values ( NOT '.')

SteveDenham — Thu, 12 Jul 2012 12:04:47 GMT

So I would assume (mmmm danger), that the same would apply for .A .B thru to .Z--the numeric missing codes?

Steve Denham

Re: Delete negative values ( NOT '.')

Haikuo — Thu, 12 Jul 2012 12:25:16 GMT

See, this is exactly the reason I would never want to miss your post. Thanks!

Haikuo

Re: Delete negative values ( NOT '.')

Haikuo — Thu, 12 Jul 2012 12:27:37 GMT

Using the information from following link to construct your comparison:

http://support.sas.com/documentation/cdl/en/lrcon/62955/HTML/default/viewer.htm#a000989180.htm

Haikuo

Re: Delete negative values ( NOT '.')

SteveDenham — Thu, 12 Jul 2012 12:31:07 GMT

A-HA!

That was the link I was looking for this morning.

So the .A thru .Z missing values are GREATER THAN .

Thanks Hai.kuo!

Steve Denham

Re: Delete negative values ( NOT '.')

data_null__ — Thu, 12 Jul 2012 12:32:58 GMT

The same what?

Re: Delete negative values ( NOT '.')

SteveDenham — Thu, 12 Jul 2012 12:38:46 GMT

We're now out of sync. See Hai.Kuo's post with the link to the order of missing values. That clarifies everything, and should satisfy the OP's question.

Since I have almost no need to use the "coded" missing values, I made the mistake of thinking that .A thru .Z were the equivalent of ._ for ordering (i.e. "the same").

Learned a lot this morning.

Steve Denham

Re: Delete negative values ( NOT '.')

data_null__ — Thu, 12 Jul 2012 13:01:10 GMT

According to this link ._ is not special missing value. I have always called it a special missing value.

Re: Delete negative values ( NOT '.')

Howles — Thu, 12 Jul 2012 13:29:53 GMT

Try

if sign(variable) EQ -1 then delete ;

See

http://www.sascommunity.org/wiki/Numeric_transformations

spraynardz90 wrote:

Hi I am looking to delete any observation with a negative value...
Generally I use:

data x; set x; if variable<0 then delete; run;

However this deletes blank observations such as '.' along with any negative values.

Does anyone know how to delete only negative values?

Unfortunately the statement below doesn't work:

data x; set x; if variable=- then delete; run;

Re: Delete negative values ( NOT '.')

Astounding — Thu, 12 Jul 2012 13:59:31 GMT

In agreement with the validity if previous posts, here is how it is usually done:

if (.Z < variable < 0) then delete;

Good luck.

Re: Delete negative values ( NOT '.')

Haikuo — Thu, 12 Jul 2012 14:19:14 GMT

Ksharp's has already provided a safe approach:

if var<0 and not missing(var) then delete;

This will include all kinds of missing values ranging from '._' to '.Z'.

Regards,

Haikuo

Re: Delete negative values ( NOT '.')

kevgermany — Thu, 12 Jul 2012 14:28:16 GMT

Your problem is that you're mixing up numeric and character data. And SAS' automatic conversions are trying to help.

By saying it deletes when variable = ' ', this means variable is character. So comparing to 0 forces an automatic conversion to numeric. (check the warnings in the log)...

So you need to use the following delete statement

if put(variable,5.) < 0 and put(variable,5.) ne . then delete.

SAS will still complain about invalid numeric data, but your tests will work correctly. Use an appropriate numeric format instead of the 5. if that's not appropriate.