Date:         Fri, 16 Jan 2009 09:55:47 +0000
Reply-To:     karma <dorjetarap@GOOGLEMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         karma <dorjetarap@GOOGLEMAIL.COM>
Subject:      Re: convert data format for longitudinal data analysis with
              discrete responses
Comments: To: Newstat <chunyuanfei@gmail.com>
In-Reply-To:  <d6a0d8f10901160151w7304698g54ac1b807f980ac2@mail.gmail.com>
Content-Type: text/plain; charset=ISO-8859-1

Sorry, just noticed we can do this with fewer variables

data want;
   array temp_month {6} _temporary_;
   do until (last.id);
      set have;
      by id;
      if disease =: "Y" then temp_month[months] = months;
      end;
   do months = 1 to 6;
      disease = ifc(missing(temp_month[months]),"No","Yes");
      temp_month[months] = .z;
      output;
      end;
run;
proc print;run;

2009/1/16 karma <dorjetarap@googlemail.com>:
> Here's a DOW solution with temporary arrays that should do it in one pass.
>
> data have;
> input ID disease$ months;
> cards;
> 100        No                .
> 101        Yes              3
> 101        Yes              5
> 103         No               .
> 104         No               .
> 105         Yes             1
> 105         Yes             3
> 105         Yes             6
> ;run;
>
> data want(drop=_:);
>   array temp_month {6} _temporary_;
>   do _i_ = 1 by 1 until (last.id);
>      set have;
>      by id;
>      if disease =: "Y" then temp_month[months] = months;
>      end;
>   do _n_=1 to 6;
>      disease = ifc(missing(temp_month[_n_]),"No","Yes");
>      months = _n_;
>      temp_month[_n_] = .z;
>      output;
>      end;
> run;
> proc print;run;
>
> 2009/1/16 Newstat <chunyuanfei@gmail.com>:
>> We follow up the children from birth until 6 months, and the outcome
>> of interest is respiratory diseases. Now my data set looks like as
>> follows (simple version).
>>
>> ID        disease         age (months) (at which the disease occured)
>> 100        No                .
>> 101        Yes              3
>> 101        Yes              5
>> 103         No               .
>> 104         No               .
>> 105         Yes             1
>> 105         Yes             3
>> 105         Yes             6
>> .....
>> .....
>> .....
>> ...
>>
>> I want to create a new data set, with each row corresponding to one
>> observation-month for one child.
>> ID        disease         age (months)
>> 100        No                1
>> 100        No                2
>> 100        No                3
>> 100        No                4
>> 100        No                5
>> 100        No                6
>>
>> 101        No                1
>> 101        No                2
>> 101        Yes              3
>> 101        No                4
>> 101        Yes               5
>> 101        No                6
>>
>> 103        No                1
>> 103        No                2
>> 103        No                3
>> 103        No                4
>> 103        No                5
>> 103        No                6
>>
>> 104        No                1
>> 104        No                2
>> 104        No                3
>> 104        No                4
>> 104        No                5
>> 104        No                6
>>
>> 105        Yes               1
>> 105        No                 2
>> 105        Yes               3
>> 105        No                 4
>> 105        No                 5
>> 105        Yes               6
>> .....
>> .....
>> .....
>> ...
>>
>> How can I convert this data set so that I can build general linear
>> mixed models? Thanks,
>> Vivian
>>
>