@Jeff_DOC wrote: Yes, I have tried removing the by group statement and the code will run. However, the use would like the percentages by the by groups included in the data. Removing the by statement allows it to run but gives me a percentage over the data set not within the groups. The suggestion was not to omit the BY statement but to omit the BY variables in the TABLES statement. This should have no impact at all on the output datasets. In particular, the percentages will not change. ... View more

Hello @Jeff_DOC,   Have you tried ballardw's suggestion to omit the BY variables in the TABLES statement? This should have no impact at all on the output datasets. But it avoids the warnings WARNING: Variable var_1 already exists on file WORK.WANT_01. WARNING: Variable var_2 already exists on file WORK.WANT_01. in the log and it simplifies the code (maybe also the internal processing of PROC FREQ) and helps avoiding typos/syntax errors such as table /*var_1*/ * var_2 * var_3 / out = want_01  noprint; ... View more

Hi @ANKH1,   According to the documentation, the default of the REF= option of the CLASS statement is LAST, which means: If the (formatted) values of CLASS variable group are '1' and '2', a hazard ratio of 0.0567 for group 1 indicates a considerably lower risk of experiencing the event in group 1 than in group 2 (e.g., event = infection, '1' = highly effective vaccine vs.  '2' = placebo). ... View more

@hhchenfx: Have you changed the requirements? I'm asking because four solutions were suggested, only three of which, applied to your sample data, produce the results in your WANT dataset (in the case of Kurt Bremser's code after adapting the start date), but you have accepted the one that ignores gaps in the sequence of dates. ... View more

Hi @hhchenfx,   Here is a DATA step solution: %let w=4; /* length of time window (in days) */ data want(drop=i j); set have; by id date; array d[0:&w] _temporary_; /* current and last four dates */ array v[0:&w] _temporary_; /* current and last four values */ if first.id then call missing(of d[*], of v[*]); d[mod(_n_,&w+1)]=date; v[mod(_n_,&w+1)]=value; max=.; do i=1 to &w; j=mod(_n_+&w+1-i,&w+1); if date-&w>d[j] then leave; if v[j]>max then max=v[j]; end; run; It assumes that the input dataset (HAVE) is sorted by ID DATE and that there are no duplicate dates within an ID. ... View more

Hello @Feyng819,   Why do you want to use a heavy tool for a simple task? For replacements of one character by another the TRANSLATE function is likely more efficient (and easier to work with) than PRXCHANGE:   Example: 31 data _null_; 32 c='102/103-Apple&Orange'; 33 c=translate(c, ' ', '/-&'); 34 put c; 35 run; 102 103 Apple Orange The first character in the third argument is replaced by the first character in the second argument (here: a space). In the absence of more characters in the second argument, the remaining characters of the third argument are replaced with spaces by default. (So, even a zero-length string could be used as the second argument because of this default behavior.) ... View more

@Taliah wrote: You wrote " To avoid a result of 1 in case of a blank line (no delimiters)" -  you mean if row 1 has no modofiers, and no data? I mean no delimiters and no data except blanks.   When I add the addition you gave and the file has modifiers and data in row 1,  it subtracts 1 from the count (as you wrote), ... No, it doesn't. It subtracts 1 "in that case," i.e., in case of a blank line. The Boolean value (_infile_=" "), which is subtracted, equals 1 only if the equality _infile_=" " is true  (this is the case of a blank line, regardless of the number of blanks it contains, zero or more) and 0 otherwise. That is, in almost all cases nothing is subtracted. ... View more

Hi @JackZ295,   In addition to the invalid date literal to be corrected, the three assignment statements must be enclosed in a DO-END block, not connected with AND. if record_id="CABA010" then do; qi100a=upcase(redcap_data_access_group); qi100b="C-ABA-DQQ-000001"; qi100d="22FEB2024"d; end;   @PaigeMiller: Blanks and various special characters are allowed in a date literal (albeit not recommended), even text following the year component. 705 data _null_; 706 d1='***1\_.~^~._/Jan:-)24==>Happy New Year!'d; 707 put d1; 708 d2='01JAN2024'd; 709 put d2; 710 run; 23376 23376 NOTE: DATA statement used (Total process time): ... View more

Hello @Rabelais,   I think in PROC SQL you might use the undocumented MONOTONIC function -- as ChanceTGardener has already suggested -- and so did I (reluctantly) when someone asked a similar question in 2019 (see Re: DO Loop with OUTPUT in SQL?). So, if you have a dataset at hand with at least N observations (e.g., the 428 obs. in SASHELP.CARS for N=100), you can use something like this: %let N=100; proc sql; create table want(where=(i<=&N)) as select monotonic() as i from sashelp.cars; quit; Otherwise it gets more complicated as you need to create such a dataset (or view) from scratch, e.g., like view BASE below: proc sql noprint; create table seed (x num); insert into seed set x=0 set x=1; select 'seed'||repeat(',seed',int(log2(&N)-1)) into :ds from seed; create view base as select ' ' from &ds; create table want(where=(i<=&N)) as select monotonic() as i from base; quit; Again, this is really not recommended. Use a DATA step instead. ... View more

Hello @Malmroth,   @Malmroth wrote: I calculate this in the main sample by first running the crude model using proc phreg and saving the output, then running the adjusted model and saving the output. Then I merge the saved output from both models and calculate the difference using the formula which I posted above. Since this requires multiple steps, I have not figured out how to calculate the bootstrap confidence intervals for this difference in hazard ratio. I think you can still perform the steps described by Rick (and in the Cassell [2010] reference linked in his blog). Both the PROC PHREG step "running the crude model" and the step "running the adjusted model" would use a BY statement (by Replicate). The saved outputs from these steps contain the BY variable (Replicate), so merging "the saved output from both models" should be straightforward. A single DATA step should then suffice to "calculate the difference" using your formula. Percentiles of these differences (one per replicate) form the lower and upper bounds of a bootstrap confidence interval. ... View more