Correct, and a warning to those who
might face this problem.
However, from Murray Lowe's reply to Antoine
Terracol (not included below) it may safely be
inferred that this is not the case here.
Nick Cox
[email protected]
>>> Seb Buechte
you are right, -sort- sorts missing _numeric_ values to the end.
Still, from what I observe in case of a string variable - sort - sorts
missings, i.e. empty strings, to the top, which certainly makes sence.
However, if "cost" was a string variable the command you have
presented will not work as wanted..
Nick Cox
> The -sort- sorts missing values to the end
> of each panel. So afterwards if any values in the panel
> are missing, then the last one will be too. That
> is necessary and sufficient information for a -drop-.
>
> The -drop- then drops all observations in the panel
> if (iff) the last one is missing.
Christian Holz
> > I think, however, that Nick's approach does not work, if a value for
> > year 5 is there and another year has a missing value, as
> > Nick's command
> > only checks the last observation of each ID group.
> > I might be wrong, but in case I am not, it's worth mentionning...
Nick Cox
> > Another way of doing this, without any new
> > > variables:
> > >
> > > bysort ID (Cost) : drop if missing(Cost[_N])
Antoine Terracol
> > >>I would try something like :
> > >>
> > >>generate tag=(cost==.)
> > >>egen toberemoved=sum(tag), by(ID)
> > >>drop if toberemoved>0
> > >>drop tag toberemoved
> > >>
> > >>
> > >>You will need to replace the "cost==." in the fisrt line by a more
> > >>general way to tag your erroneous values (such as "cost==. |
> > >>cost>9999")
Murray Lowe
> > >>>I am working with a large dataset and have discovered that
> > >>
> > >>some of the data
> > >>
> > >>>are missing values or have erroneous values. The data is
> > >>
> > >>panel data with
> > >>
> > >>>observations per individual over a 5 year period. For example:
> > >>>
> > >>>ID Year Cost
> > >>>
> > >>>1 1 100
> > >>>1 2 200
> > >>>1 3 500
> > >>>1 4 150
> > >>>1 5 x
> > >>>2 1 100
> > >>>2 2 200
> > >>>2 3 500
> > >>>2 4 600
> > >>>2 5 100
> > >>>
> > >>>The problem is this: If an individual has a missing /
> > >>
> > >>erroneous value for a
> > >>
> > >>>particular year, I want to exclude ALL of their
> > >>
> > >>observations from the
> > >>
> > >>>dataset. In the example patient 1 would be removed from the dataset
> > >>>entirely. How can this be done through an automated-type process?
> > >>>Essentially I need a code / method that looks for the
> > >>
> > >>anomalous data;
> > >>
> > >>>identifies the patient and then removes all of their
> > >>
> > >>observations from the dataset.
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
