Date:         Sun, 17 Jun 2007 22:34:13 -0700
Reply-To:     David L Cassell <davidlcassell@MSN.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         David L Cassell <davidlcassell@MSN.COM>
Subject:      Re: Randomly picking a procedurecode for each doctor
In-Reply-To:  <7367b4e20706150829k500fbdeah5d957dccce2bb2ed@mail.gmail.com>
Content-Type: text/plain; format=flowed

datanull@GMAIL.COM sagely replied:
>
>I think, you want to use STRATA DOCID and N=1.  If I understand correctly.
>
>data work.docs;
>   input docid:$6. procedurecode:$5. @@;
>   cards;
>001111   21740 001111   21740 001111   21740 001111   21740 001111   21740
>001111   20680 001111   20680 001112   50240 001112   50240 001112   50240
>001112   52601 001112   52601 001112   55845 001113   48140 001113   48140
>001113   48140 001113   48140 001113   48140 001113   48150 001113   48150
>001113   47130 001113   47135 001114   53500 001114   53500 001114   53500
>001114   53450 001114   53450 001115   21045 001115   21045 001115   21045
>001115   21040 001115   21040
>;;;;
>   run;
>proc surveyselect
>      seed=20062001
>      data=docs
>      method=SRS
>      n=1
>      out=surgerycase1;
>   strata docid;
>   run;
>proc print;
>   run;
>
>
>On 6/15/07, Annie Lee <hummingbird10111@hotmail.com> wrote:
>>Hi,
>>
>>I have a data set with doctor's id (multiple records for each doctor) and
>>procedurecode.
>>
>>This time, I would like to pick one randomly selected procedurecode for
>>each (unique) doctorid in order to avoid any bias in selecting
>>procedurecode.
>>
>>I tried doing this using proc surveyselect.
>>one question I have regarding using proc surveyselect-- is there a way not
>>to specify the sampsize in advance?
>>In the future, I will have different number of records every quarter in
>>this data set and I do not want to calculate the unique number of docid
>>beforehand to assign the sampsize manually.
>>
>>I would appreciate any help.  Thank you.  -Eunice
>>
>>
>>
>>proc surveyselect data = surgerycase  method = SRS rep = 1
>>                       sampsize = ?? out = surgerycase1;
>>id _all_;
>>run;
>>
>>
>>
>>Results I would like to have:
>>
>>docid   procedurecode (it is randomly picked so it could be any
>>procedurecode)
>>
>>001111   21740
>>001112   50240
>>001113   48140
>>001114   53500
>>001115   21045
>>
>>data set:
>>
>>docid   procedurecode
>>
>>001111   21740
>>001111   21740
>>001111   21740
>>001111   21740
>>001111   21740
>>001111   20680
>>001111   20680
>>001112   50240
>>001112   50240
>>001112   50240
>>001112   52601
>>001112   52601
>>001112   55845
>>001113   48140
>>001113   48140
>>001113   48140
>>001113   48140
>>001113   48140
>>001113   48150
>>001113   48150
>>001113   47130
>>001113   47135
>>001114   53500
>>001114   53500
>>001114   53500
>>001114   53450
>>001114   53450
>>001115   21045
>>001115   21045
>>001115   21045
>>001115   21040
>>001115   21040
>>

D0 has supplied the solution I had in mind.

I just want to add one thing.  Because of his construction, the input
data set is *already* sorted on DOCID, the stratum variable.  If your
data are not yet sorted (or indexed) on DOCID, you would need to
sort/index on this variable first.  Otherwise, the proc will complain.
A lot.

When doing stratified sampling, PROC SURVEYSELECT uses the
N= option to tell how many records to pick in each stratum, rather
than for the whole data set.  So N=1 is exactly what you asked for.
It may *not* be what you should really be thinking about...

HTH,
David
--
David L. Cassell
mathematical statistician
Design Pathways
3115 NW Norwood Pl.
Corvallis OR 97330

_________________________________________________________________
Don’t miss your chance to WIN $10,000 and other great prizes from Microsoft
Office Live http://clk.atdmt.com/MRT/go/aub0540003042mrt/direct/01/