Hi Reeza, I used a very dumb way to fix the problem. I don't think think that my approach will be helpful for the future. Since I am very new to SAS (just study SAS for several weeks), it's very difficult for me to understand what you said that why I think of doing like following.
I just keep everything as is in the Event data (include all holidays & weekends) but .with longer time period (-200, +4) to cover for the weekends & holidays). Then I merged the event data with the daily stock return data. Since daily stock return data was from CRPS database, it only includes trading days, when I merged, the holidays & weekends automatically disappeared in the merged file.
I splited this merged data into "estper" table including all observation before the event date and into "evntper" table including all observation after the event date. In table "estper" I do some minor calculation for the abnormal return to have a table named "mmparam" and use this table to merge with the "eventper" to create "ar" tables.
In the table "evntper", variable "before" that I mentioned in a post related with this post, only includes a value of 0, 3 or 4. I then looked at the eventdate to see how many dates are on Friday and its following Mondays are holidays and noted these specific dates, then I create three tables. One table includes Friday-event date with normal working day on Monday. Second table includes Friday-event date with holiday on Monday. Final table includes all other event dates. After that, I manually change the value of "before" from 3 to 1 (Monday is the third day of the Event; Sat=1st and Sun=2nd were deleted when I merged 2 table). For Friday-event date with Monday is a holiday, I change the value of "before" from 4 to 1 (Sat=1, Sun=2, Mon=3 are not working/trading day).
After I successfully changed the value of the variable "before", I concatenated the 3 tables to have the final clean table "arclean".
Quite a long story =))
BTW, how to master my SAS skill? I feel it too difficult 😞
data Event_intermediate;
format date EventDate yymmddn8.;
set stock.events (keep=EventDate Ticker);
do Date = EventDate - 200 to EventDate + 4;
output;
end;
run;
proc sort data=Event_intermediate;
by TICKER Date;
run;
data stock.want;
merge
Event_intermediate (in=a)
stock.StockDayClean (in=b)
;
by TICKER Date;
if a and b;
before=Date-EventDate;
Ewday=weekday(EventDate);
run;
proc sort data=stock.want;
by TICKER EventDate Date;
run;
data stock.estper stock.evntper;
set stock.want;
if before<0 then output stock.estper;
if before>=0 then output stock.evntper;
run;
proc reg data=stock.estper outest=mmparam (rename=(intercept=alpha MrkReturn=beta)
keep=TICKER EventDate Date intercept MrkReturn) noprint;
by TICKER EventDate;
model return = MrkReturn;
quit;
data ar;
merge stock.evntper mmparam;
by TICKER EventDate;
ar = return - alpha - beta*MrkReturn;
run;
data arED6H arED6WK arEDR;
set ar;
if EventDate in ('31AUG2012'd, '12FEB2016'd) then output arED6H;
else if Ewday=6 then output arED6WK;
else output arEDR;
run;
data arED6Hclean;
set arED6H;
if WdayStock=3 then before=1;
run;
data arED6WKclean;
set arED6WK;
where before<4;
if WdayStock=2 then before=1;
run;
data arEDRclean;
set arEDR;
where before<2;
run;
data stock.arclean;
set arEDRclean arED6WKclean arED6Hclean;
run;
proc sort data=stock.arclean;
by ticker EventDate;
run;
Click image to register for webinar
