topic Re: Timeseries data - subsetting in SAS Programming

Timeseries data - subsetting

petterco — Wed, 23 Apr 2014 16:12:52 GMT

Hi.

I have a small problem with creating a dataset I need for an analysis.

Lets assume I have an ID variable "serial", and two different string variables: - "period" in the form 201403

- "period_renewed" in the same form

First what I want to do is to delete all data where period>period_renewed.

Further I want to keep only the observations from the last 6 periods from the point in time where period=period_renewed.

Someone out there might have a simple solution to this problem, but I'm completely blank as I'm quite new to using SAS.

Any feedback would be extremely welcome.

/petter

Re: Timeseries data - subsetting

Michelle — Tue, 20 May 2014 21:50:00 GMT

the following will create a dataset called subset that excludes anything where period>period_renewed:

data subset;

set original;

if period>period_renewed then delete;

run;

alternatively, you could use:

data subset;

set original;

where period<=period_renewed;

run;

not really sure what you are asking for in the second part.

Re: Timeseries data - subsetting

PGStats — Wed, 21 May 2014 01:45:28 GMT

Assuming there are no duplicates :

proc sort data=have(where=( period<=period_renewed)) out=temp; by serial descending period; run;

data want;

i = 0;

do until (last.serial);

set temp; by serial;

i + 1;

if i <= 6 then output;

end;

drop i;

run;

proc sort data=want; by serial period; run;

(Untested)

Re: Timeseries data - subsetting

PGStats — Wed, 21 May 2014 03:19:23 GMT

Or perhaps, simpler:

proc sort data=have; by serial period; run;

data want;

retain firstSerial;

set temp; by serial;

if first.serial then firstSerial = _n_;

if period>=period_renewed then do;

do point = max(firstSerial, _n_ - 5) to _n_;

set temp point=point;

output;

end;

firstSerial = constant('BIG');

end;

drop point firstSerial;

run;

(Untested)

Re: Timeseries data - subsetting

petterco — Wed, 21 May 2014 11:19:41 GMT

Thank you both for your suggestions
I solved the problem through changing format of the string to a date variable via the input function,

and then used the intck function + an extra variable to select the right interval.

@PGstats I learned something new from you! Thanks

/Petter