Re: st: RE: perform regression on minimum number of observations stata

[Nahla Betelmal <nahlaib@gmail.com>]

Perfect, it works great. Many thanks indeed.

On 29 July 2013 17:53, Nick Cox <njcoxstata@gmail.com> wrote:
> This is not difficult.
>
> Evidently, you need to check that all the variables you want to
> include in the regression are non-missing. So, extend the -count- to
> include that.
>
> count if industry == `i' & year == `y' & !missing(DV, IV_1, IV_2, IV_3)
> Nick
> njcoxstata@gmail.com
>
>
> On 29 July 2013 17:43, Nahla Betelmal <nahlaib@gmail.com> wrote:
>> Thanks Nick and Gergorio, the loop works, but still the number of
>> observations run by the regression  less than 15 in some cases. for
>> example see this output
>>
>>
>> year = 2011 and industry= 9
>>    20
>>
>>       Source |       SS       df       MS
>> Number of obs =      13
>> -------------+------------------------------
>>   F(  3,     9) =    5.14
>>        Model |  .159024959     3   .05300832             Prob > F      =  0.0242
>>     Residual |  .092789059     9  .010309895            R-squared     =  0.6315
>> -------------+------------------------------
>>    Adj R-squared =  0.5087
>>        Total |  .251814017    12  .020984501           Root MSE      =  .10154
>>
>> ---------------------------------------------------------------------------------
>> DV |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
>> ----------------+----------------------------------------------------------------
>>  IV_1 |  -.6177928   .3150983    -1.96   0.082    -1.330595     .095009
>> IV_2 |   .6304449   .2551211     2.47   0.036     .0533208    1.207569
>>  IV_3 |  -.0369961   .0173076    -2.14   0.061    -.0761486    .0021564
>>   _cons |  -.0884111   .0439023    -2.01   0.075    -.1877251    .0109029
>> ---------------------------------------------------------------------------------
>>
>> Although the number of year-industry observations here is 20, only 13
>> had enough data for the regression. Is there a way to make the number
>> of observations regressed  at least 15.
>>
>> I totally understand if there is no way to do it, your help has been great.
>>
>> Thanks a million for the help and time.
>>
>> Nahla
>>
>>
>> On 29 July 2013 17:12, Impavido, Gregorio <GImpavido@imf.org> wrote:
>>> Try changing
>>>
>>> count if industry == `i' & `year' == `y'
>>>
>>> with
>>>
>>> count if industry == `i' & year == `y'
>>>
>>> -----Original Message-----
>>> From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Nahla Betelmal
>>> Sent: Monday, July 29, 2013 11:46 AM
>>> To: statalist@hsphsun2.harvard.edu
>>> Subject: Re: st: RE: perform regression on minimum number of observations stata
>>>
>>> Thanks Nick, but I got an error with the loop
>>>
>>> year = 1989 and industry= 1
>>> ==1989 invalid name
>>> r(198);
>>>
>>> I  added `  ' for industry in the line:  count if industry == `i' & `year' == `y'. I still get the same error but with alteration
>>>
>>> year = 1989 and industry= 1
>>> ==1 invalid name
>>> r(198);
>>>
>>>
>>> Could it be that I do not have enough observations at year 1989 and industry 1
>>>
>>> Thanks again, I highly appreciate your time. Also thanks for the advice about the magic number and regression model.
>>>
>>> Nahla
>>>
>>>
>>> On 29 July 2013 16:24, Nick Cox <njcoxstata@gmail.com> wrote:
>>>> Sorry for previous incomplete reply.
>>>>
>>>> Nick
>>>> njcoxstata@gmail.com
>>>>
>>>> On 29 July 2013 16:20, Nick Cox <njcoxstata@gmail.com> wrote:
>>>>
>>>> Gregorio's loop can be modified something like this
>>>>
>>>> forval y=1989/2012 {
>>>>        forval  i= 1/57 {
>>>>         di "year = `y' and industry = `i'"
>>>>         count if industry == `i' & `year' == `y'
>>>>         if r(N) > 15 {
>>>>         reg DV IV_1 IN_2 IN_3 if  Industry== `i' & year==`y'
>>>>         }
>>>> }
>>>>
>>>> Whatever magic numbers or rules of thumb you read about,
>>>>
>>>> 1. Don't take them too literally.
>>>>
>>>> 2. Increase the desirable number according to how many parameters you
>>>> are estimating.
>>>>
>>>> 3. Use sensible models. I don't usually expect pure linear regressions
>>>> to work well with firm-year data.
>>>>
>>>> Nick
>>>> njcoxstata@gmail.com
>>>> *
>>>> *   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/
>>> *
>>> *   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/
>>>
>>> *
>>> *   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/
>> *
>> *   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/
> *
> *   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/
*
*   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/