So in what you said previously "within" countries and "within" years had
quite different meanings.
What you want looks more like
gen byte good = !missing(year, Country, var) & (var > 0)
bysort good Country : egen rank_var = rank(var)
Nick
[email protected]
[email protected]
this is an example of what I mean....
Within var1 I wish to rank countries (I have provided an example of 2
below) on that specific variable within that year.
Thank you, and once again, thanks so much for your attention to this...
VAR1 VAR2
Belize 1995 298700000.00 15.764373
Belize 1996 308400000.00 17.196754
Belize 1997 312300000.00 11.364923
Belize 1998 331100000.00 5.5330906
Belize 1999 365400000.00 -0.29874134
Belize 2000 399100000.00 -6.1305733
Belize 2001 433500000.00 -11.962405
Belize 2002 446900000.00 -17.794237
Belize 2003 457600000.00 -23.626069
Belize 2004 540800000.00 -29.457901
Belize 2005 574800000.00 -35.289733
Belize 2006 598700000.00 -41.121565
Brazil 1980 110900000000.00 -0.64295287
Brazil 1981 124500000000.00 -0.5691551
Brazil 1982 133200000000.00 -0.49535733
Brazil 1983 95930000000.00 -0.42155956
Brazil 1984 98600000000.00 -0.34776179
Brazil 1985 105000000000.00 -0.27396402
Brazil 1986 125700000000.00 -0.20016625
Brazil 1987 138200000000.00 -0.12636848
Brazil 1988 155900000000.00 -0.05257071
Brazil 1989 199700000000.00 0.02122706
From: Nick Cox <[email protected]>
> Your ranks will not exceed 1 if you have at most one observation
> for each combination of country and year, regardless of
> validity.
>
> I understood from your postings that you seek rankings within
> cross-combinations of country and year. That would be futile if
> your data are defined uniquely by country and year.
>
> You should show us a segment of your data if this is not clear.
> Do bear in mind that we rely entirely on what you say.
>
> [email protected]
>
> Thanks Nick, this is now excluding the values less than 0.
> However, it appears to be assigning all the positive values as
> "1" as opposed to ranking them in order "1,2,3,4...." Am I doing
> something wrong?
>
> ----- Original Message -----
> From: Nick Cox <[email protected]>
> > I don't know what "does not seem to work" means in your case,
> > but one way forward is something like this.
> >
> > First, segregate observations that are good and bad for your
> > purpose. Let's say "good" observations are non-missing on
> these
> > variables and positive on -var-.
> >
> > gen byte good = !missing(year, Country, var) & (var > 0)
> >
> > Second,
> >
> > bysort good Country year : egen rank_var = rank(var)
> >
> > Third, use only
> >
> > ... if good
> > [email protected]
> >
> > Further to my query below:
> > I just found that I have a number of missing and negative
> > values. Is there anyway I can exclude these cases when I
> assign
> > the country ranks?
> >
> > [email protected]
> >
> > > I am trying to assign ranks to country observations within
> > > years. By this I mean that I want to assign a score for each
> > > observation (within a country) within that specific year.
> > > So far, I have calculated:
> > > egen rank_var = rank(var) ---- this seems to provide ranks
> to
> > > countries, but does not take into account yearly changes.
> > > I tried:
> > > bysort year Count: egen rank_var= rank(var)--- but this does
> > not
> > > seem to work either. Is this something to do with
> > subscripting?
*
* 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/
