Ignoring inappropriate values is not unreasonable, provided that it
calls your attention to what it has done. The accuracy of (pre-2003)
LINEST is comparable to PROC GLM in SAS. Excel gets slammed and SAS
doesn't because SAS warns the user when results are not numerically
trustworthy.

Jerry

Myrna Larson wrote:

Hi, Jerry:

I was surprised by the OP's statement that "the analysis program ... just
ignores all that". If it just throws out the data, the results will be
worthless. If it in fact interprets those entries by calculating a date, the
user should be aware of that.

But the bottom line is that there should be data validation in place that
disallows entries that aren't dd.mm.yy; and after all of the fuss about Y2K, 2
digit years should have been disallowed too.



On Mon, 09 May 2005 14:46:11 -0400, "Jerry W. Lewis"
wrote:


Regardless of how they are formatted, Excel dates are stored as the
number of days since the beginning of 1900. Provided that the entry is
an Excel date or can be coerced into an Excel date, the formula should work.

Data QC is often the biggest portion of data analysis.

Jerry

Steve Wylie wrote:


Yeah, that's the trouble. The dates have not been inputted consistently.
There are many false entries where people have put "16 Dec" and no year (it
should all be dd.mm.yy) or just "age 42" or rubbish like that. The

analysis

program I use just ignores all that, whereas Excel throws up an error.

And I suspect your formula doesn't like the years in two-digit format
either...

Thanks, but I did it using the analysis program in the end. Shame tho.

Steve