windorize without replacing missing data

Waqar — Wed, 19 Jun 2013 12:50:11 GMT

I need to windorize data. I am using the following code to do so. However, it replaces the missing values too.

proc univariate data=acc_data noprint;

var cash CF Lev Q DVC AT B2M LagAT;

output out=_winsor pctlpts=1 99 pctlpre=__cash __CF __Lev __Q __DVC __AT __B2M __LagAT;

run;

data acc_data_win (drop=_:);

set acc_data;

if _n_=1 then set _winsor;

array wlo {*} __cash1 __CF1 __Lev1 __Q1 __DVC1 __AT1 __B2M1 __LagAT1;

array whi {*} __cash99 __CF99 __Lev99 __Q99 __DVC99 __AT99 __B2M99 __LagAT99;

array wval {*} wcash wCF wLev wQ wDVC wAT wB2M wLagAT;

array val {*} cash CF Lev Q DVC AT B2M LagAT;

do _V=1 to dim(val);

end;

run;

Any suggestion to avoid replacing missing values and windorize only the non missing values.

Thanks in advance.

Regards,

Waqar

Re: windorize without replacing missing data

PGStats — Wed, 19 Jun 2013 14:53:30 GMT

To avoid replacing missing values, you must exclude missing values from the lower range (they rank as minus infinity) :

do _V=1 to dim(val);

if . < val{_V} < wlo{_V} then val{_V} = wlo{_V};

else if val{_V} > whi{_V} then val{_V} = whi{_V};

end;

you might need a retain statement on the _winsor variables.