R help - Jan 2012 - Checking dates for entry errors

Hello Everyone,
?
I have a question about how best to check dates for?entry errors. I recently
discovered that R will read?the incorrectly entered date "11/23/21931"
without producing a warning or an?error message at least under some
circumstances.
?
> as.Date("11/23/21931", format = "%m/%d/%Y")
[1] "2193-11-23"
> as.Date("21931-11-23")
Error in charToDate(x) : 
  character string is not in a standard unambiguous format

Similarly, under some circumstances, R will convert an impossible date like
February 31, 2011 to NA rather than issuing a warning.
> as.Date("02/31/2011", format = "%m/%d/%Y")
[1] NA
> as.Date("2011-02-31")
Error in charToDate(x) : 
? character string is not in a standard unambiguous format
?
In the former case, one could easily lose the date rather than recognizing it is
in error and needs to be corrected.
?
So my question is how best to check these sorts of date values.
?
So far, I've been checking date values with things like:
?
sort( unique(DOB) )
sort( unique(substr(DOB, 1, 4) ) )
sort( unique(substr(DOB, 6, 7) ) )
sort( unique(substr(DOB, 9, 10) ) )
?
These are good for seeing, say, year values that are clearly in error, but
don't do much to catch the impossible date I cited above.
?
How can I use R to better scrutinize my date data? Is there any way to make it
complain more when there is a problem with my date data?
?
Thanks,
?
Paul
?
?

On 1/11/2012 11:07 PM, Paul Miller wrote:
> Hello Everyone,
>
> I have a question about how best to check dates for entry errors.
Try using regular expression matching and the functions grep, strsplit, 
regexpr etc.
If you are not familiar with regex: bit a bumpy road of getting into it 
but in the long term definitely worth the effort !

cheers

On 1/11/2012 11:07 PM, Paul Miller wrote:
> Hello Everyone,
>
> I have a question about how best to check dates for entry errors.
It depends on the study.
  1. Simple: use as.Date on the data, and find resultant missing values.
You might have to add a format to the as.Date call: if I use
as.Date("2001-02-29", format="%Y-%m-%d") I get an NA as
desired (no Feb
29th that year); without it a message that the function can't guess the
right format.

  2. Better: with data set checks.  Say that I have dates of birth,
study enrollment, study events (lab tests), and termination.
    Age at entry within the study eligibity criteria
    All study events happen between erollment and termination
    Termination after enrollment
    etc