Calculating best response - data step logic issue

knveraraju91 — Wed, 26 Sep 2018 14:47:59 GMT

Dear,

I have the following sample data from which i need to calculate best response and date of confirmation(DOC). Please help. Thank you

I calculate best response and DOC by following

1. If the response is CR or PR at visits, then I calculate number of days between visits(use "date" variable values). The number of days ge 28 than best response is earlier visit response and earlier visit date as DOC

eg: for id=1 number of days between visit1 and visit2 = 60(greater than 28) so best response="CR" and DOC=2017-01-01

I need to calculate between all visits

for id=1 number of days between visit1 and visit3 = 90(greater than 28) so best response="CR" and date of confirmation=2017-01-01
for id=1 number of days between visit2 and visit3 = 30(greater than 28) so best response="CR" and date of confirmation=2017-03-01

at the end of datastep for id=1:

id bestresponse DOC
1 CR 2017-01-01
1 CR 2017-01-01
1 CR 2017-03-01

for id=2, number of days between visit1 and visit2 = 10(less than 28) so final response="SD" and date of confirmation=blank
for id=2, number of days between visit1 and visit3 = 40(greater than 28) so final response="CR" and date of confirmation=2017-05-20. So on for next visits

id finalresponse DOC
2 SD
2 CR 2017-05-20
2 CR 2017-05-20
2 CR 2017-05-20
2 CR 2017-05-30
2 CR 2017-05-30
2 CR 2017-07-01

for id=5, after "CR" there is NE , so i need to take difference between "randomdate" and "date"(not between two dates). if (2017-01-01-2016-10-30) greater than 42 than final response=SD, else NE. DOC is blank

id final response DOC
5 SD

for id=7, 1.first visit=CR and second visit=NE so i take difference between date and radte.if greater than 42 than final response=SD, else NE. DOC is blank. For this subject i need to ignore second visit and third vist as second visit is NE.
2. (first visit -third visit)the number of "NE" between two CR is one. if there is one NE then i need to take difference between first CR and second CR. if gt than 28 than CR, else SD

id finalresponse DOC
7 SD
7 CR 2017-01-01

for id =8, 1.first visit=CR and second visit=NE so i take difference between date and radte.if greater than 42 than final response=SD, else NE. DOC is blank. For this subject i need to ignore second visit and third vist as second visit and third are NE.
2. (first visit -third visit)the number of "NE" between two CR response is two. if there is more than one NE then the final reponse= SD. I should not take difference between first and last CR

id final response DOC

8 SD

for ID=9, 1.1.first visit=PR and second visit=NE so i take difference between date and radte.if greater than 42 than final response=SD, else NE.
2. here more than one NE between two PR. When the response is PR, i take difference between first PR and last PR. If gt 28 han final reponse=PR. else SD

id finalresponse DOC
9 SD

9 PR 2017-01-01

For id =10

id finalresponse DOC
10 PR 2017-01-01 (first and second visit)
10 PR 2017-01-01 (first and third visit)
10 PR 2017-01-01 (first and fourth visit)
10 PR 2017-01-01 (first and fifth visit)
10 PR 2017-03-01 (second and third visit)
10 PR 2017-03-01 (second and fourth visit)
10 PR 2017-03-01 (second and fifth visit)
10 PR 2017-04-01 (third and fourth visit)
10 PR 2017-04-01 (third and fifth visit)
10 CR 2017-05-20 (third and fifth visit)

The intermediate output dataset i calculated is below. The finalresponse values are formated as CR=1, PR=2, SD=3 and PD=5,NE=4. Then I sorted the dataset by id finalresponse doc and took first,date to get output I need. My code working fine but lengthy and not getting the output i need when the new data is added.

Intermediate dataset
id fianlresponse DOC
1 CR 2017-01-01
1 CR 2017-01-01
1 CR 2017-03-01
2 SD
2 CR 2017-05-20
2 CR 2017-05-20
2 CR 2017-05-20
2 CR 2017-05-30
2 CR 2017-05-30
2 CR 2017-07-01
5 SD
7 SD
7 CR 2017-01-01
8 SD
9 SD
9 PR 2017-01-01
10 PR 2017-01-01
10 PR 2017-01-01
10 PR 2017-01-01
10 PR 2017-01-01
10 PR 2017-03-01
10 PR 2017-03-01
10 PR 2017-03-01
10 PR 2017-04-01
10 PR 2017-04-01
10 CR 2017-05-20

output i needed:

id final response doc
1 CR 2017-01-01
2 CR 2017-05-20
5 SD
7 CR 2017-01-01
8 SD
9 PR 2017-01-01
10 CR 2017-05-20

proc format;
value $aval
    'CR' = 1
    'PR' = 2
    'SD' = 3
    'PD' = 5
    'NE' = 4	
    ;
data one;
input id date$ 4-13 +1 response$ +1 randomdate $18-27;
datalines;
01 2017-01-01 CR 2016-10-30     
01 2017-03-01 CR 2016-10-30
01 2017-04-01 CR 2016-10-30      
02 2017-05-20 CR 2016-10-30
02 2017-05-30 CR 2016-10-30
02 2017-07-01 CR 2016-10-30
02 2017-08-30 CR 2016-10-30                             
05 2017-01-01 CR 2016-10-30
05 2017-01-30 NE 2016-10-30                             
07 2017-01-01 CR 2016-10-30
07 2017-01-30 NE 2016-10-30   
07 2017-02-21 CR 2016-10-30
08 2017-01-01 CR 2016-10-30
08 2017-01-30 NE 2016-10-30    
08 2017-02-30 NE 2016-10-30
08 2017-04-01 CR 2016-10-30
09 2017-01-01 PR 2016-10-30
09 2017-01-30 NE 2016-10-30    
09 2017-02-30 NE 2016-10-30
09 2017-04-01 PR 2016-10-30
10 2017-01-01 PR 2016-10-30     
10 2017-03-01 PR 2016-10-30
10 2017-04-01 PR 2016-10-30      
10 2017-05-20 CR 2016-10-30
10 2017-06-20 CR 2016-10-30
;

Re: Calculating best response - data step logic issue

RW9 — Wed, 26 Sep 2018 15:03:47 GMT

This is a Q&A board, I do not see a question here. Seems like you have a reasonable spec there how are you going with it? Or were you after a contractor to do some work for you?

topic Re: Calculating best response - data step logic issue in SAS Programming

Calculating best response - data step logic issue

Re: Calculating best response - data step logic issue