Re: st: "First numlist in weights( ...) invalid" error when using tssmooth to calculate a simple moving average

[Nick Cox <njcoxstata@gmail.com>]

Also, my -egen- function -filter()- in -egenmore- (SSC) will I think
do exactly what you want.

Nick

On Thu, Sep 6, 2012 at 6:59 PM, Nick Cox <njcoxstata@gmail.com> wrote:
> Sorry; I evidently didn't make myself clear enough, or you've missed
> some very minor ironies of style. Where the error arises is not in
> doubt.
>
> -set trace- is sufficient to tell exactly where the error occurs and
> using it lies behind my previous post. Also, the error occurs when
> int(N/2) exceeds 1600, which as you say is when N >= 3202.
>
> I haven't copied my post to Stata tech support, so you may want to
> press them gently on the point.
>
> Nick
>
> On Thu, Sep 6, 2012 at 6:42 PM, Aaron Kirkman <ak1795mailserv@gmail.com> wrote:
>> I don't know if this code is sufficient to figure out exactly where
>> the error occurs, but errors occur when _N >= 3202 on my system (Stata
>> MP 12.1, Unix).
>>
>> ##
>> set more off
>> forvalues i = 3100/3300 {
>>         clear
>>         capture set obs `i'
>>         gen x = runiform()
>>         gen t  = _n
>>         capture tsset t
>>         capture tssmooth ma wtma = x, weights(1 2 3 4 <5>)
>>         if _rc == 198 {
>>                 display as error `i'
>>         }
>> }
>> ##
>>
>> Aaron
>>
>> On Thu, Sep 6, 2012 at 7:56 AM, Nick Cox <njcoxstata@gmail.com> wrote:
>>> Yes; I can reproduce this. My impressions are
>>>
>>> 1. Your syntax is legal so far as -tssmooth- documentation is concerned.
>>>
>>> 2. -tssmooth- is passing your numlist to -numlist- for checking, but
>>> there is a conflict of views, which leads to an error message.
>>>
>>> -tssmooth- is saying that as far as it is concerned, the numlist can
>>> be 2674 elements long (in essence, half the number of your
>>> observations). However, -numlist- won't play with candidate numlists
>>> longer than 1600 elements, so it bails out without trying to examine
>>> the list.
>>>
>>> There is, I think, confusion between two distinct tests, which must be
>>> both be passed. -tssmooth- should check before it passes your numlist
>>> to -numlist- that the numlist is OK as far as the time series aspect
>>> is concerned. Then the call to -numlist- should not specify a maximum.
>>>
>>> A wild guess is that -numlist- was developed on series less than about
>>> 3200 values long, where this doesn't bite.
>>>
>>> My tentative three-letter summary is "bug". StataCorp will have a view.
>>>
>>> Nick
>>>
>>> On Thu, Sep 6, 2012 at 1:11 PM, Aaron Kirkman <ak1795mailserv@gmail.com> wrote:
>>>> I noticed an error in the code I copied to the mailing list, although
>>>> it's merely a typo and fixing it doesn't correct the -numlist- error.
>>>> The -copy- line should not have a semicolon, and instead should be
>>>> this:
>>>>
>>>> copy "http://ichart.finance.yahoo.com/table.csv?s=PNRA&a=05&b=10&c=1991&d=07&e=24&f=2012&g=d&ignore=.csv";
>>>> pnra.csv, replace
>>>>
>>>> I can reproduce this error on both Unix and Windows machines (which
>>>> may not be relevant), but can anyone else reproduce this error with
>>>> the code in the previous thread?
>>>>
>>>> Thank you,
>>>> Aaron Kirkman
>>>>
>>>> On Wed, Sep 5, 2012 at 8:32 PM, Aaron Kirkman <ak1795mailserv@gmail.com> wrote:
>>>>> Dear Statalist,
>>>>>
>>>>> I'm using the -tssmooth- command to calculate a simple moving average
>>>>> of stock prices, but I get a syntax error of "First numlist in
>>>>> weights(1 2 3 4 <5>) invalid" when running these commands.
>>>>>
>>>>> ##
>>>>>
>>>>> copy "http://ichart.finance.yahoo.com/table.csv?s=PNRA&a=05&b=10&c=1991&d=07&e=24&f=2012&g=d&ignore=.csv";
>>>>> pnra.csv, replace
>>>>> insheet using pnra.csv, comma clear
>>>>> gen dateInt = date(date, "YMD")
>>>>> drop date
>>>>> rename dateInt date
>>>>> tsset date
>>>>> tssmooth ma wtma1 = adjclose, weights(1 2 3 4 <5>)
>>>>>
>>>>> ##
>>>>>
>>>>> I also receive an error if I specify other variations of this numlist
>>>>> or other weights entirely, e.g.
>>>>>
>>>>> tssmooth ma wtma2 = adjclose, weights(1/4 <5>)
>>>>> tssmooth ma wtma3 = adjclose, weights(1(1)4 <5>)
>>>>>
>>>>>
>>>>> Specifying a window does work, however.
>>>>>
>>>>> tssmooth ma wtma4 = adjclose, window(4 1)
>>>>>
>>>>> However, this syntax works for another data set, so I'm not quite sure
>>>>> about the problem with this code. For example, this code works when
>>>>> using a dataset from Baum's "Introduction to Modern Econometrics Using
>>>>> Stata."
>>>>>
>>>>> ##
>>>>>
>>>>> use http://www.stata-press.com/data/imeus/grunfeld, clear
>>>>> tsset
>>>>> tssmooth ma wtma1 = invest, weights(1 2 3 4 <5>)
>>>>> tssmooth ma wtma2 = invest, weights(1/4 <5>)
>>>>> tssmooth ma wtma3 = invest, weights(1(1)4 <5>)
>>>>>
>>>>> ##
>>>>>
>>>>> Is there a nuance to the syntax that I'm missing?
>>>>>
*
*   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/