Data Structure
in [SAS]
|
Prev: SAS capabilities for determining covariate coefficients
Next: Call for Paper: International Journal of Mathematics and Computation (IJMC).
From: Al on 10 May 2010 15:41 All: This is the data i have .. pat vis res 1 1 12 1 1.1 13 1 1.2 12 1 2 11 2 1 09 2 2 14 2 3 11 3 1 10 3 2 11 3 3 1.1 3 4 10 The desired data structure is one record per pat.. i able to do it using proc transpose but i want to know how this is can be done using arrays.. Any suggestions .. Thanks in advance Al
From: Patrick on 11 May 2010 07:52 Why code some data step when Proc Transpose does it so easily. That's the real power of SAS. In order to use arrays you would have to pass the data set twice AND to sort the data set previously. The first pass through the data (sorted) is to to find the biggest numer of observations per by group (pat) which then is the number of array elements, the second time to assign the values to the array elements - and then: if last.pat then output;
From: Al on 11 May 2010 10:05 On May 11, 6:52 am, Patrick <patrick.mat...(a)gmx.ch> wrote: > Why code some data step when Proc Transpose does it so easily. That's > the real power of SAS. > > In order to use arrays you would have to pass the data set twice AND > to sort the data set previously. > The first pass through the data (sorted) is to to find the biggest > numer of observations per by group (pat) which then is the number of > array elements, the second time to assign the values to the array > elements - and then: if last.pat then output; Just want to know how the datastep works for this kind of scenario Thanks al
From: Al on 11 May 2010 11:32 On May 11, 9:05 am, Al <ali6...(a)gmail.com> wrote: > On May 11, 6:52 am, Patrick <patrick.mat...(a)gmx.ch> wrote: > > > Why code some data step when Proc Transpose does it so easily. That's > > the real power of SAS. > > > In order to use arrays you would have to pass the data set twice AND > > to sort the data set previously. > > The first pass through the data (sorted) is to to find the biggest > > numer of observations per by group (pat) which then is the number of > > array elements, the second time to assign the values to the array > > elements - and then: if last.pat then output; > > Just want to know how the datastep works for this kind of scenario > > Thanks > al Can any one help me with code for data step using Array for the above scenario Thanks in Advance
From: Reeza on 11 May 2010 11:55
On May 11, 8:32 am, Al <ali6...(a)gmail.com> wrote: > On May 11, 9:05 am, Al <ali6...(a)gmail.com> wrote: > > > > > > > On May 11, 6:52 am, Patrick <patrick.mat...(a)gmx.ch> wrote: > > > > Why code some data step when Proc Transpose does it so easily. That's > > > the real power of SAS. > > > > In order to use arrays you would have to pass the data set twice AND > > > to sort the data set previously. > > > The first pass through the data (sorted) is to to find the biggest > > > numer of observations per by group (pat) which then is the number of > > > array elements, the second time to assign the values to the array > > > elements - and then: if last.pat then output; > > > Just want to know how the datastep works for this kind of scenario > > > Thanks > > al > > Can any one help me with code for data step using Array for the above > scenario > > Thanks in Advance- Hide quoted text - > > - Show quoted text - As stated above, you need to do one pass through the data and figure out how many records per patient, the maximum size of the array. Array sizes need to be declared at the beginning and cannot grow (last time I checked) so this method becomes fairly inefficient due to this restrictions. Else sort the data by patient, vist and use first/last with an array. Some untested code below: proc sort data=have; by pat visit; data want; set have; by pat; array pat_visits{30}; (i'm assuming 30 is the maximum, you'd have to find out. retain index=1; *what record are you dealing with for the patient; pat_visits[index]=res; *assigns value to array; if pat.last then do; *if last record for patient output and reset index; index=0; output; end; index=index+1; *increment index for next row; run; |