Re: st: Puzzling error with merge

[David Hoaglin <dchoaglin@gmail.com>]

Phil (and Joe),

Thanks for the suggestions.

I received an Excel file containing two sheets and imported them into
separate Stata data files.  Those are the files that I was trying to
merge, treating the file from Sheet 1 as the master.  On each of those
Stata data files, I used -tabulate- with StudyID as a categorical
variable (its type is string).  I expected the output to show a
frequency of 2 (or more) for any duplicate, but each value of StudyID
had a frequency of 1 in both files, and the total frequency was
correct in both files.

I'll look for extra blank observations at the end.  Multiple blank
values of StudyID would certainly show up as duplicates in the
-merge-.  If the observations are completely blank, I'm not sure how
to recognize them.  Perhaps the numerical variables in the files will
appear as missing.

David Hoaglin

On Mon, Jan 13, 2014 at 11:47 AM, Phil Schumm <pschumm@uchicago.edu> wrote:
> On Jan 13, 2014, at 8:04 AM, David Hoaglin <dchoaglin@gmail.com> wrote:
>> My key variable has no missing values in either file.
>
>
> Is that statement based on what you know to be true from the original data (i.e., as you scan through them in the Excel file), or based on the results of
>
>     assert !mi(StudyID)
>
> It is often the case that data imported from Excel contain extra (blank) observations at the end.  I don't mean to question the veracity of your statement, but given what you've said, the existence of some extra observations in the master file with missing values for StudyID appears to be the only possible explanation (unless Joe is right, and you accidentally skipped over a duplicate in the output from -tabulate-).
>
>
> -- Phil

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/