Hi @victor1893,   Here's an algorithm using an arbitrary number of months (macro variable n), &n>1, as the length of the time window: proc sort data=data(drop=col_want) out=have; by id time; run; %let n=6; /* length of time window in months */ data want(drop=_:); array _c[0:%eval(&n-2)] _temporary_; set have; by id; _prev=lag(col_have); if first.id then call missing(of _c[*]); else do; _m+1; _c[mod(_m-1,%eval(&n-1))]=ifn(_prev=0 & col_have=1,1,.); end; col_want=n(of _c[*]); run; /* Optional: */ proc sort data=want; by id descending time; run; It still assumes that the values of variable TIME are a contiguous series of months for each ID, so that counting observations is equivalent to counting months. ... View more

Hi @mmohotsi,   A simple solution is to add a STYLEATTRS statement to the PROC SGPLOT step and to specify the group colors in the DATACONTRASTCOLORS= option.   Example: proc sgplot data=Tested3; styleattrs datacontrastcolors=(lime gray purple skyblue); vline Dates / response=Amount group=sampling_point groupdisplay=overlay ; ... See Color-Naming Schemes for the various ways to specify colors (e.g., by RGB values). ... View more

Hello @alepage,   I would also prefer the format approach suggested by ballardw (where the definition 'PQ' = 'QC' in the VALUE statement should be sufficient), especially if you need the same on-the-fly replacement ('PQ' → 'QC') in several places of your programs or if there are more pairs of values.   Otherwise, yes, you can use a CASE expression as you have suggested.   Example: data have1; input prov $ x; cards; AB 1 ON 2 PQ 3 BC 4 ; data have2; input province $ y; cards; MB 555 QC 333 BC 444 ON 222 ; %let varname1=prov; proc sql; create table want as select b.province, a.x, b.y from have1 a inner join have2 b on case when a.&varname1='PQ' then 'QC' else a.&varname1 end = b.province; quit; Or use the IFC function: on ifc(a.&varname1='PQ','QC',a.&varname1) = b.province; ... View more

Hello @Manj,   You can define a format which displays 10 as "10^1", 100 as "10^2" and so on ... data fmt; retain fmtname 'expfmt'; length label $6; do e=0 to 15; start=10**e; if e=0 then label='1'; else label=cats('10^',e); output; end; run; proc format cntlin=fmt; run; ... and then add the VALUESFORMAT= option to your YAXIS statement, referring to that format: valuesformat=expfmt.   An alternative format definition could use superscripts for the exponents (10¹, 10², ...): proc format; value supexpfmt 1='1' 1e1="10(*ESC*){unicode '00B9'x}" 1e2="10(*ESC*){unicode '00B2'x}" 1e3="10(*ESC*){unicode '00B3'x}" 1e4="10(*ESC*){unicode '2074'x}" 1e5="10(*ESC*){unicode '2075'x}" 1e6="10(*ESC*){unicode '2076'x}" 1e7="10(*ESC*){unicode '2077'x}" 1e8="10(*ESC*){unicode '2078'x}" 1e9="10(*ESC*){unicode '2079'x}" 1e10="10(*ESC*){unicode '00B9'x}(*ESC*){unicode '2070'x}" 1e11="10(*ESC*){unicode '00B9'x}(*ESC*){unicode '00B9'x}" 1e12="10(*ESC*){unicode '00B9'x}(*ESC*){unicode '00B2'x}" 1e13="10(*ESC*){unicode '00B9'x}(*ESC*){unicode '00B3'x}" 1e14="10(*ESC*){unicode '00B9'x}(*ESC*){unicode '2074'x}" 1e15="10(*ESC*){unicode '00B9'x}(*ESC*){unicode '2075'x}" ; run; However, this would require a suitable font: valueattrs=(size=8 family='Arial Unicode MS') valuesformat=supexpfmt. ... View more