Date:         Thu, 20 May 2010 11:05:47 -0400
Reply-To:     Joe Whitehurst <joewhitehurst@GMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Joe Whitehurst <joewhitehurst@GMAIL.COM>
Subject:      Re: How to delete observations with all variables missing?
Comments: To: Yuewei Liu <elek.liu@gmail.com>
In-Reply-To:  <AANLkTilv038pPRkMRMGROnBS722QKatyI2Q7V5rn1gh9@mail.gmail.com>
Content-Type: text/plain; charset=ISO-8859-1

Of course, it you wanted the data in a SAS dataset, you would still have to
define the variables.  But you could avoid defining the variables for
records with all missing values.

data test;
 infile test
 input;
 if ^missing(_infile_) then input /* variable specifications*/;
run;


On Thu, May 20, 2010 at 10:58 AM, Joe Whitehurst <joewhitehurst@gmail.com>wrote:

> You can still avoid moving the data into the PDV like this:
>
>  filename test 'e:\weblog_analysis\test.txt';
> 2
> 3    data test;
> 4    infile test;
> 5    input;
> 6    if ^missing(_infile_);
> 7    run;
>
> NOTE: The infile TEST is:
>       Filename=e:\weblog_analysis\test.txt,
>       RECFM=V,LRECL=256,File Size (bytes)=85,
>       Last Modified=20May2010:10:52:08,
>       Create Time=20May2010:10:52:08
>
> NOTE: 5 records were read from the infile TEST.
>       The minimum record length was 0.
>       The maximum record length was 26.
> NOTE: The data set WORK.TEST has 3 observations and 0 variables.
> NOTE: DATA statement used (Total process time):
>       real time           0.03 seconds
>       cpu time            0.03 seconds
>
> Test.txt is a text file with 5 records with two them having all missing
> data.
>
>
> On Thu, May 20, 2010 at 8:36 AM, Yuewei Liu <elek.liu@gmail.com> wrote:
>
>> The question is, we often got some raw data set with missing observations
>> for all variables. So we have to do this. Anyway, I think the coalesce and
>> coalescec functions are nice to handle this.
>> Thank you for your reply.
>>
>> Yuewei
>>
>>
>> On Thu, May 20, 2010 at 8:12 PM, Joe Whitehurst <joewhitehurst@gmail.com>wrote:
>>
>>> I'm curious.  Wouldn't it be move efficient to put either of these
>>> functions in a where clause to prevent observations with all missing values
>>> for being loaded into the PDV at all?
>>>
>>>
>>> On Thu, May 20, 2010 at 7:06 AM, Yuewei Liu <elek.liu@gmail.com> wrote:
>>>
>>>> Data _null_, very cool...  Thank you very much.
>>>>
>>>> Yuewei
>>>>
>>>> On Thu, May 20, 2010 at 6:21 PM, Data _null_; <iebupdte@gmail.com>
>>>> wrote:
>>>>
>>>> > if missing(cats(of _all_)) then delete;
>>>> > You will need the system option MISSING=' ';
>>>> >
>>>> > or
>>>> >
>>>> >   if missing(coalesce(of _numeric_))
>>>> >      and missing(coalesceC(of _character_)) then delete;
>>>> >
>>>> > On 5/20/10, Yuewei Liu <elek.liu@gmail.com> wrote:
>>>> > > Hi all. How to delete the observation, if all values of all the
>>>> variables
>>>> > > are missing? I did it using the Array statements. Do you have some
>>>> better
>>>> > > ideas?
>>>> > > Any suggestions are appreciated.
>>>> > >
>>>> > > data have;
>>>> > > input x y;
>>>> > > length z $ 8;
>>>> > > z=ifc(y=.," ",strip(put(y,best.)));
>>>> > > cards;
>>>> > > 1 2
>>>> > > 3 1
>>>> > > 4 2
>>>> > > 3 .
>>>> > > 3 3
>>>> > > . 2
>>>> > > . .
>>>> > > 3 .
>>>> > > . .
>>>> > > . 2
>>>> > > . .
>>>> > > ;
>>>> > > data want(drop=tag i j);
>>>> > >  set have;
>>>> > > array num [*] _numeric_;
>>>> > > array char[*] _character_;
>>>> > >  tag=0;
>>>> > > do i=1 to dim(num);
>>>> > > if num[i]^=. then do;
>>>> > >  tag=1;
>>>> > > return;
>>>> > > end;
>>>> > >  end;
>>>> > > do j=1 to dim(char);
>>>> > > if char[j]^=" " then do;
>>>> > >  tag=1;
>>>> > > return;
>>>> > > end;
>>>> > >  end;
>>>> > > if tag=1;
>>>> > > run;
>>>> > >
>>>> > > Yuewei
>>>> > >
>>>> >
>>>>
>>>
>>>
>>
>