Fair enough. Could you please inform me on how to recreate the data set which I have destroyed? I would appreciate that.
First of all you do not need to run an entirely new data step to do a single calculation. Each repeated use of the
Data setname;
set setname;
construct increases changes of problem causing.
with some care all of this code could be in one data step
DATA paper.ceo_firm;
length ticker $5;
merge paper.CSRP_annual_returns2 (in=in1) paper.compustat_execucomp4 (in=in2);
by ticker;
if in1 and in2;
run;
data paper.ceo_firm;
set paper.ceo_firm;
if annual_return=. then
delete;
run;
data paper.ceo_firm;
set paper.ceo_firm (rename=(BONUS=_BONUS STOCK_AWARDS=_STOCK_AWARDS));
BONUS=input(_BONUS, ?? 8.);
STOCK_AWARDS=input(_STOCK_AWARDS, ?? 8.);
run;
*create dummy variable that indicates whether the firm's market value increases in a given year;
data paper.ceo_firm;
set paper.ceo_firm;
by ticker year;
last_return=lag(annual_return);
if not first.ticker then
increase=(annual_return > last_return);
else
increase=.;
run;
data paper.ceo_firm;
set paper.ceo_firm;
if gender='MALE' then male=1; else male=0;
run;
data paper.ceo_firm;
set paper.ceo_firm;
if salary=0 then log_salary=0;
else log_salary=log(salary);
run;
*create variable log_annual_return;
data paper.ceo_firm;
set paper.ceo_firm;
if annual_return=0 then log_annual_return=0;
else log_annual_return=log(annual_return);
run;
*create variable log_bonus;
data paper.ceo_firm;
set paper.ceo_firm;
if bonus=0 then log_bonus=0;
else log_bonus=log(bonus);
run;
*create variable log_TOTAL_SEC;
data paper.ceo_firm;
set paper.ceo_firm;
if TOTAL_SEC=0 then log_TOTAL_SEC=0;
else log_TOTAL_SEC=log(TOTAL_SEC);
run;
You do similar things with other data sets. If you find that you need an additional variable then go back to the first data step that references the data you need and insert the code there. Then you can find all of the likely problems in one place instead of hunting through a dozen different data steps.
Also note that you have this step in there twice:
data paper.ceo_firm;
set paper.ceo_firm (rename=(BONUS=_BONUS STOCK_AWARDS=_STOCK_AWARDS));
BONUS=input(_BONUS, ?? 8.);
STOCK_AWARDS=input(_STOCK_AWARDS, ?? 8.);
run;
Which is likely to cause at least some warnings about SAS converting numeric to string values the second time around.
I suspect some of your issues may come from use of proc import to begin with. Read the values as needed with a data step at the beginning.
Proc import for CSV will create a data step code that appears in the log. I almost always only use Proc Import to generate that data step and then modify to ensure the variable types are correct (I see several places where it appears you make numeric from imported character variables, the data step could address that), of the desired length, assign labels and formats I want. I also will add needed variables, data cleaning code (deletions, spelling corrections sometimes, expected value checks such as dates in a range of years).
The data step also means that if I get a similar CSV in the future I can read it by changing the infile name and create a different output data set name if desired and all of the variables will be the same: name, label, type and formats. Which is unlikely with certain data sources relying on Proc Import.
And if there are two or more related files to read then I can insure that the common variables all have the same name, length and type so combining data works much easier.
Without knowing your report logic I would be concerned about the way you are deleting the multiple ceo records. Is your intent to exclude from analysis any year completely that had the ceo change?
Click image to register for webinar
