Something like the following should work: data have (drop=position length); informat x y $30.; input x & y &; if scan(y,1) eq 'th' then do; call scan(x, FINDW(x, 'th',' ', 'e')-1, position, length); z=substr(x,position); end; else z=y; cards; 34 45 th avenue th avenue South west 34 street street ; Art, CEO, AnalystFinder.com ... View more

Not an area I'm familiar with, but doesn't Rho in R simply control the type of test used (i.e., with 0=logrank and 1=peto)?   If so, you're already controlling it with the test option.   Art, CEO, AnalystFinder.com   ... View more

@drjorg : As I mentioned, one can only guess unless you provide some example records that closely approximate your actual data.   However, that said, there was a logical error in the way you modified the code, and I added the probably necessary 'i' modifier.   See if the following comes closer to what you want/expected:   data limit; informat id 8.; informat Class_I_p $1.; informat Result_Date date9.; format Result_Date date9.; informat comments $255.; input ID Class_I_p RESULT_DATE comments &; cards; 1 1 02feb2019 On Monday I had 1.5 scoops of vanilla ICE cream, Tuesday we had 2 scoops of chocolate Ice cream, one day we also had 3 scoops of strawberry icE cream, but I can't remember when. 2 1 02feb2019 Over the week I had 4 scoops of raspberry ICE cream, 3 scoops of vanilla Ice cream 3 1 02feb2019 On 1/20/2019, I had 3 scoops of chocolate ice cream. 4 0 02feb2019 On 2/1/2019 I didn't want any ice cream 5 1 02feb2019 On 2/1/2019 I had 1 scoop of lemon iCe ; data try6; set limit(keep=ID Class_I_p RESULT_DATE comments); i=0; j=0; ixp=0; ixs=-2; if Class_I_p=1 then do; do while (findw(comments,"ICE",' ,','ei',ixs+3)); i+1; if i eq 1 then ixs+3; ixp=findw(comments,"ICE",' ,','ei',ixs)+ixp; j+1; SCOOPS=catx(' ',scan(comments,ixp-4,' '),scan(comments,ixp-3,' ')); Flavor = scan(comments,ixp-1,' '); output; ixs=findw(comments,"ice",' ,','i',ixs)+3; end; end; run; %transpose(data=try6,out=want, var=scoops flavor, id=j, by=id RESULT_DATE) Art, CEO, AnalystFinder.com   ... View more

@DR_Jorg:   You'd have to provide some sample data that reflects your actual data. My first guess is that you have to add the 'i' modifier to the findw calls to ignore the case of the string 'ice'. I.e., to capture any variation of capitalization. However, I'd have to run your code against actual data to identify why you are getting the numeric to character conversion note.   Art, CEO, AnalystFinder.com   ... View more

If searching for 'ice' rather than 'ice cream' would suffice, then here is one method: data have (keep=id j scoops flavor); informat comments $255.; length scoops flavor $10; input id comments &; i=0; j=0; ixp=0; ixs=-2; do while (findw(comments,"ice",' ,','e',ixs+3)); i+1; if i eq 1 then ixs+3; ixp=findw(comments,"ice",' ,','e',ixs)+ixp; if (scan(comments,ixp-3,' ') eq 'scoops' or scan(comments,ixp-3,' ') eq 'scoop') and scan(comments,ixp-2,' ') eq 'of' then do; j+1; scoops=catx(' ',scan(comments,ixp-4,' '),scan(comments,ixp-3,' ')); flavor = scan(comments,ixp-1,' '); output; end; ixs=findw(comments,"ice",ixs)+3; end; cards; 1 On Monday I had 1.5 scoops of vanilla ice cream, Tuesday we had 2 scoops of chocolate ice cream, one day we also had 3 scoops of strawberry ice cream, but I can't remember when. 2 Over the week I had 4 scoops of raspberry ice cream, 3 scoops of vanilla ice cream 3 On 1/20/2019, I had 3 scoops of chocolate ice cream. 4 On 2/1/2019 I didn't want any ice cream 5 On 2/1/2019 I didn't want any ice cream, but the day before I had 1 scoop of lemon ice ; %transpose(data=have,out=want, var=scoops flavor, id=j, by=id) /* uses transpose macro which can be downloaded from http://www.sascommunity.org/wiki/A_Better_Way_to_Flip_(Transpose)_a_SAS_Dataset */ Art, CEO, AnalystFinder.com   ... View more