Re: st: reverse lookup

[Nick Cox <njcoxstata@gmail.com>]

That way lies -levelsof-.

Nick

On Wed, Jan 9, 2013 at 2:54 PM, Jeph Herrin <stata@spandrel.net> wrote:
> Yes, the sort works as well for strings.
>
> I'm thinking of a program will return the list of matches, along with a flag
> regarding uniqueness, because in some contexts it's good have the set of
> matching values. But no need to post it here (grin).
>
> cheers,
> Jeph
>
>
> On 1/8/2013 10:19 PM, Nick Cox wrote:
>>
>> The approach in my earlier code could be extended.
>>
>> program vallookup, sort
>>         version 8.2
>>         syntax varname [if] [in] [, local(str) scalar(str) ]
>>
>>         marksample touse, strok
>>         qui count if `touse'
>>         local nuse = r(N)
>>         if `nuse' == 0 error 2000
>>
>>         capture confirm numeric variable `varlist'
>>
>>         if _rc == 0 {
>>                 su `varlist' if `touse', meanonly
>>                 if r(min) != r(max) {
>>                         di as err "specification not satisfied by single
>> value"
>>                         exit 498
>>                 }
>>
>>                 di r(min)
>>                 if "`local'" != "" {
>>                         c_local `local' = r(min)
>>                 }
>>                 if "`scalar'" != "" {
>>                         scalar `scalar' = r(min)
>>                 }
>>         }
>>         else {
>>                 sort `touse' `varlist'
>>                 if `varlist'[_N - `nuse' + 1] != `varlist'[_N] {
>>                         di as err "specification not satisfied by single
>> value"
>>                         exit 498
>>                 }
>>
>>                 di `varlist'[_N]
>>                 if "`local'" != "" {
>>                         c_local `local' = `varlist'[_N]
>>                 }
>>                 if "`scalar'" != "" {
>>                         scalar `scalar' = `varlist'[_N]
>>                 }
>>         }
>> end
>>
>>
>> On Tue, Jan 8, 2013 at 10:58 PM, Jeph Herrin <stata@spandrel.net> wrote:
>>>
>>> String variables are a problem all their own. I usually do something
>>> like:
>>>
>>>   encode strvar, gen(strvar_coded)
>>>   sum strvar_coded if period==1
>>>   local rate1= ///
>>>      cond(`r(min)'==`r(max)',`=: label (strvar_coded) `=r(min)'',"")
>>>
>>> which however can run into trouble if there are too many values to
>>> -strvar-.
>>>
>>>
>>>
>>> On 1/8/2013 4:48 PM, Nick Cox wrote:
>>>>
>>>>
>>>> Here's a sketch. (Also, what about string variables?)
>>>>
>>>> program vallookup
>>>>          version 8.2
>>>>          syntax varname(numeric) [if] [in] [, local(str) scalar(str) ]
>>>>
>>>>          marksample touse, strok
>>>>          qui count if `touse'
>>>>          if r(N) == 0 error 2000
>>>>
>>>>          capture confirm numeric variable `varlist'
>>>>
>>>>          su `varlist' if `touse', meanonly
>>>>          if r(min) != r(max) {
>>>>                  di as err "specification not satisfied by single value"
>>>>                  exit 498
>>>>          }
>>>>
>>>>          di r(min)
>>>>          if "`local'" != "" {
>>>>                  c_local `local' = r(min)
>>>>          }
>>>>          if "`scalar'" != "" {
>>>>                  scalar `scalar' = r(min)
>>>>          }
>>>> end
>>>>
>>>> On Tue, Jan 8, 2013 at 9:33 PM, Jeph Herrin <stata@spandrel.net> wrote:
>>>>>
>>>>>
>>>>> Yes, the Mata construct is the ideal. And obviously, one must have 1-1
>>>>> mapping; this I usually check by:
>>>>>
>>>>>    sum rate if period==1
>>>>>    local rate=cond(`r(min)'==`r(max)',r(min),.)
>>>>>
>>>>> I was thinking of writing some programs to do lookups like this, since
>>>>> I
>>>>> have been doing so many, and thought I'd ask first for an alternative.
>>>>>
>>>>> thanks,
>>>>> Jeph
>>>>>
>>>>>
>>>>> On 1/8/2013 2:27 PM, Nick Cox wrote:
>>>>>>
>>>>>>
>>>>>>
>>>>>> My short answer is that yes, this is awkward, but you are working with
>>>>>> the most obvious way to do it in Stata. The problem is that in general
>>>>>>
>>>>>> ... if <condition>
>>>>>>
>>>>>> is not guaranteed to identify precisely one observation. It might
>>>>>> yield one, or zero or more than one.
>>>>>>
>>>>>> In your case you need == in your code and can use
>>>>>>
>>>>>> su rate if period == 1, meanonly
>>>>>> local value = r(min)
>>>>>>
>>>>>> The misnamed -meanonly- is quieter and more efficient. If the
>>>>>> condition identifies precisely one observation, then clearly r(min),
>>>>>> r(mean), r(max) will be identical.
>>>>>>
>>>>>> The problem is discussed from a different angle in
>>>>>>
>>>>>> SJ-6-4  dm0025  . . . . . . . . . .  Stata tip 36: Which observations?
>>>>>> Erratum
>>>>>>            . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
>>>>>> N.
>>>>>> J. Cox
>>>>>>            Q4/06   SJ 6(4):596                              (no
>>>>>> commands)
>>>>>>            correction of example code for Stata tip 36
>>>>>>
>>>>>> SJ-6-3  dm0025  . . . . . . . . . . . . . .  Stata tip 36: Which
>>>>>> observations?
>>>>>>            . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
>>>>>> N.
>>>>>> J. Cox
>>>>>>            Q3/06   SJ 6(3):430--432
>>>>>> (no
>>>>>> commands)
>>>>>>            tip for identifying which observations satisfy some
>>>>>>            specified condition
>>>>>>
>>>>>> Mata is not surprisingly less awkward here:
>>>>>>
>>>>>> : y = 1::10
>>>>>>
>>>>>> : x = runiform(10,1)
>>>>>>
>>>>>> : x , y
>>>>>>                      1             2
>>>>>>         +-----------------------------+
>>>>>>       1 |  .5044846558             1  |
>>>>>>       2 |  .0174561641             2  |
>>>>>>       3 |   .680281796             3  |
>>>>>>       4 |  .9221656218             4  |
>>>>>>       5 |  .1094441491             5  |
>>>>>>       6 |  .7122591983             6  |
>>>>>>       7 |   .765775156             7  |
>>>>>>       8 |  .0226029507             8  |
>>>>>>       9 |  .9540165765             9  |
>>>>>>      10 |  .2686450339            10  |
>>>>>>         +-----------------------------+
>>>>>>
>>>>>> : select(x, y :== 1)
>>>>>>      .5044846558
>>>>>>
>>>>>> Nick
>>>>>>
>>>>>> On Tue, Jan 8, 2013 at 7:07 PM, Jeph Herrin <stata@spandrel.net>
>>>>>> wrote:
>>>>>>
>>>>>>> I've just written the same awkward code for the untoldth time, and
>>>>>>> I'm
>>>>>>> thinking there must be a better way to do it.
>>>>>>>
>>>>>>> The problem is to get a particular value of a variable into a local
>>>>>>> which
>>>>>>> corresponds to a particular value of another variable. I think this
>>>>>>> is
>>>>>>> usally call reverse lookup. For example, I might have -period- and
>>>>>>> -rate-
>>>>>>> and want to store the value of -rate- which corresponds to period =
>>>>>>> 1.
>>>>>>> My
>>>>>>> lazy solution is
>>>>>>>
>>>>>>>
>>>>>>>     sum rate if period = 1
>>>>>>>     local rate1 `=r(mean)'
>>>>>>>
>>>>>>> That is, I summarize a single observation, then put the mean in
>>>>>>> local.
>>>>>>> Is
>>>>>>> there a better way to do this?
*
*   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/