Re: st: RE: RE: -encode- with predetermined sorting

[Nick Cox <njcoxstata@gmail.com>]

My point 2 was about your original example. From the details here the
flavour of your problem is a bit clearer, and now I advise checking
out -labmask- (SJ) and also

SJ-8-2  gr0034  . . . . . . . . . .  Speaking Stata: Between tables and graphs
        (help labmask, seqvar if installed) . . . . . . . . . . . .  N. J. Cox
        Q2/08   SJ 8(2):269--289
        outlines techniques for producing table-like graphs

and also -sencode- (SJ, SSC),

-encode- is an official command. It's not usually difficult to write
your own alternative if it doesn't do what you want.

Nick

On Tue, Jun 12, 2012 at 3:52 PM, Impavido, Gregorio <GImpavido@imf.org> wrote:
> Thank you Nick.  Your suggestions prompted me to do a little more homework: hence, my belated response.
>
> My original question is about both the points you raise (I think).
>
> Regarding your point 2., my loop below produces such labels in advance following the sorting of a. But in 2. below you suggest that this could be done with a one line command.  Are you thinking of some use of -levelsof-? Because if I have (say) many values in b, a one line of the type -label define lblname 1 "a" 4 "b" ... - is rather tedious (although it forces the idiosyncratic sorting I want).  But most importantly it is not readily changeable if I want to change the sort order (BTW: The whole use of this is to produce -graph twoyay bar- charts where the xlabels are sorted in any arbitrary fashion; contrary to -graph bar-, there is no -sort()- option in -graph twoway bar- (that I could find...))
>
> Regarding your point 1., I managed to force -enocode- to use the arbitrary sorting produced in advance (by the loop below) by coding -encode a, gen(c2) label(lblname)-.  This produces the same variable c produced by -label values c lblname- below. Then, I am more confused: is there a concrete advantage then in using -encode- afterall if I have to specify the label name in advance?  Wouldn't be more useful if encode had a -sort(varlist)- option?

Nick Cox

> 1. The value labels for an idiosyncratic -encode- need to exist in advance. Idiosyncratic just means here an encoding that -encode- would not produce by default.
>
> 2. In your case you could define them with a single command line, which would be quicker to type and to execute than your code.
>
> I am guessing that your question is more about the first than the second.

Impavido, Gregorio

> I have the following dataset:
>
> . list
>
>      +--------+
>      | a    b |
>      |--------|
>   1. | 10   a |
>   2. | 40   b |
>   3. | 30   c |
>   4. | 20   d |
>      +--------+
>
> If I encode b and generate c, I get:
>
> . list *, nolabel
>
>      +-------------+
>      | a    b   c  |
>      |-------------|
>   1. | 10   a   1  |
>   2. | 40   b   2  |
>   3. | 30   c   3  |
>   4. | 20   d   4  |
>      +-------------+
>
> Now, I want to have a different sort order for c, say as a function of a:
>
> . list *, nolabel
>
>      +-------------+
>      | a    b   c  |
>      |-------------|
>   1. | 10   a   1  |
>   2. | 40   b   4  |
>   3. | 30   c   3  |
>   4. | 20   d   2  |
>      +-------------+
>
> Is there a more rapid way than the following to obtain what I want?
>
> a
> gen c = _n
> forvalues i=1/4 {
>        label define lblname `i' "`: di b[`i']'", add
>        }
> label values c lblname

*
*   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/