* Hi,

The dataset DATES contains valid and not valid date values, and the code in the two following data sets try to identify which ones are valid and which ones are not. The rules are: 1) '0000000000' is not a valid date value. 2) 01.02.02 is a valid date, even though we don't know whether it is 1902 or 2002. 3) The date must exist in the calender (f.ex, not 40.10.1998). 4) The value can be blank. ;

data dates; attrib date_values length=$12.; infile cards; input @1 date_values $12.; datalines; 0000000000 01.05.2001 31.12.2002

01.02.02 31.09.1999 40.10.1998 20.04.1800 22644840 01.01.21000 01.01.19900 01.01.19899 01.01.19999 01.01.20000 01.01.20100 ; run;

*The following data set seems to solve the problem OK, don't you think ?;

data not_valid1 (drop = datevalue) is_valid1 (drop = datevalue) ; set dates; datevalue=input(compress(date_values,'.'), ddmmyy10.); if date_values ne '' and datevalue = . then output not_valid1; else output is_valid1; run;

*But this one is not doing so well, as far as I can see from the output. What is wrong with this ?; data not_valid2 is_valid2 ; set dates; if (date_values ne ' ' and date_values ne '0000000000') then do; if missing(input(date_values,??yymmdd8.)) then output not_valid2; end; else output is_valid2; run;

*Is there any other techniques to validate date values that are character variables ?

Regards, Rune ;