Re: st: match string variables

[Nick Cox <njcoxstata@gmail.com>]

I don't think these problems ever have clean, simple solutions. You
just need to keep plugging away until -merge- works.

One specific [pun intended] detail is to standardise binomials
[taxonomic sense] of the form Genusname speciesname so that
speciesname always begins with lower case, as in Homo sapiens, Homo
sasuser and Homo statauser.

Standardising to that in your main dataset is

gen taxon2 = upper(substr(taxon, 1,1)) + lower(substr(taxon, 2, .))

Another standardisation is to generic and specific names only in your
species list as in

gen taxon2 = word(taxon,1) + " " + word(taxon, 2)

In other words, you can work at the corresponding variables in each
dataset until you have a good chance of a successful -merge-.

-merge- does not depend on files having the same number of observations!

More general advice is contained in

SJ-8-3  dm0039  . . .  Stata tip 64: Cleaning up user-entered string variables
        . . . . . . . . . . . . . . . . . . . . . . . .  J. Herrin and E. Poen
        Q3/08   SJ 8(3):444--445                                 (no commands)
        tip on how to clean up user-entered string variables

Nick

P.S. 1 taxon, 2 taxa.

On Wed, Apr 13, 2011 at 10:18 PM, "Lukas Bösch" <L.Boesch@gmx.de> wrote:
>
> I am working with the CITES trade data and my aim is to analyze the export of 130 countries from 1990 to 2009 with a logistic model. CITES regulates the international trade in endangered species. The export data for Afghanistan, for example, looks like this:
>
> year       taxon          term       unit        country         value
>
> 1990      Falco Cherrug   live         -            AF              0
> 1991      Falco Cherrug   live         -            AF              0
> 1992      Falco Cherrug   live         -            AF              0
> 1993      Falco Cherrug   live         -            AF              0
>
> In the case of Afghanistan, the data contains 180 rows, with nine different taxon. In some cases it contains up to 8000 rows with 2000 taxon. I know that I could also have shown the data in a wide form with much fewer rows...
>
> Now I want to create a variable “indigenous” with 1 if the exported taxon exists in the country or 0 if not. In order to get this I copied the species lists for all 130 countries from the CITES homepage, which looks like this (again for Afghanistan):
>
> Accipiter badius (Gmelin, 1788)
> Accipiter gentilis (Linnaeus, 1758)
> Accipiter nisus (Linnaeus, 1758)
> Acinonyx jubatus (Schreber, 1775)
> Acipenser nudiventris Lovetzky, 1828
>
> This list contains 131 taxon and I sorted it out in order to get rid of the years, the commas and so on.
>
> Accipiter badius
> Accipiter gentilis
> Accipiter nisus
> Acinonyx jubatus
> Acipenser nudiventris Lovetzky
>
> I have tried different variations of merge and joinby, I looked at the ado files _gsoundex, nearmrg, nmatch and reclink but I haven’t been able to create the “indigenous” variable so far.
> There are two major problems. The first one is that the taxon in the species list doesn’t always match exactly with the taxon in the export data. For example, Falco cherrug, in the export data is listed as Falco Cherrug Gray in the species list. The second problem is that the species list has a different number of observations than the export data and they dont fit logically together.
> I need something like: If the taxon from the export data is on the species list, then “indigenous” = 1, if the taxon from the export data is not on the species list, then “indigenous” = 0.

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