Re: st: How to get mean coefficients and t-statistics from several regressions

[Nick Cox <njcoxstata@gmail.com>]

In a strong sense you are both right and wrong.

-regress- returns e(r2_a) and many, probably most, regression commands
are based on -regress-. Often somebody writes a wrapper program that
does some preparation and some processing before and after a call to
-regress-.

But (a very big but) what's crucial is exactly what a wrapper program
returns. If it does not pass on a e-class result, that result is lost.
Nick
njcoxstata@gmail.com


On 30 July 2013 16:25, Nahla Betelmal <nahlaib@gmail.com> wrote:
> Thanks for the clarification, and sorry about that. I genuinely
> assumed that every regression command return  e(r2_a).
>
> Sorry and thanks again
>
> Nahla
>
> On 30 July 2013 16:11, Nick Cox <njcoxstata@gmail.com> wrote:
>> -xtfmb- (SSC) does not return -e(r2_a)-.
>>
>> To find that out takes two minutes running an example and checking the code.
>>
>> A Stata expert is just someone who just gets into certain habits:
>> reading the help, looking at the code and running examples to check
>> what is going on.
>>
>> I guess you need to clone the program and modify the cloned version to
>> work out adjusted R-square.
>>
>> Nick
>> njcoxstata@gmail.com
>>
>>
>> On 30 July 2013 16:00, Nahla Betelmal <nahlaib@gmail.com> wrote:
>>> Thanks Nick for the reply, I tried that as well, but it gave me
>>> missing adjusted R-squared, although the value of R-squared was
>>> available. So I assumed that something wrong with this command,
>>> because the papers reported adjusted R-square.
>>>
>>> Why the adjusted R square is missing? It would be great if there is a
>>> way to get it.
>>>
>>> Also, I wonder if there is another way than statsby to get the mean of
>>> the regressions betas and mean adjusted R square and matched standard
>>> errors.
>>>
>>> Many thanks again
>>>
>>> Nahla
>>>
>>> On 30 July 2013 15:09, Nick Cox <njcoxstata@gmail.com> wrote:
>>>> Quite where the error is coming from is for you to tell us.
>>>>
>>>> Meanwhile this seems confused:
>>>>
>>>>  forval  i= 1/50 {
>>>>        statsby _b e(r2)  e(r2_a) , by(industry):xtfmb DV IV1 IV2 IV3 if
>>>>  industry== `i'
>>>> }
>>>>
>>>> Much of the point of the -statsby- command is that it does the looping
>>>> for you. It seems more likely that you want
>>>>
>>>> statsby _b e(r2)  e(r2_a) , by(industry): xtfmb DV IV1 IV2 IV3
>>>>
>>>> Nick
>>>> njcoxstata@gmail.com
>>>>
>>>>
>>>> On 30 July 2013 14:43, Nahla Betelmal <nahlaib@gmail.com> wrote:
>>>>> Hi Richard, thanks for your previous comments, and sorry for the delay
>>>>> in replying. I had to do some readings to see how researchers solve
>>>>> this issue.
>>>>>
>>>>> In industry-year context rather than firm-year context, authors seem
>>>>> to report mean of the mean rather than just Fama-MacBeth  time series
>>>>> mean.
>>>>>
>>>>> first they perform Fama-MacBeth on each industry alone to get the time
>>>>> series mean . In other words, they perform cross-sectional  on the
>>>>> firms-years observations in each industry to produce  time series mean
>>>>> of each industry. This could be easily obtained by the following
>>>>> Fama-MacBeth command available in Stata:
>>>>>
>>>>> tsset firm_id year
>>>>>
>>>>> forval  i= 1/50 {
>>>>>     xtfmb DV IV_1 IV_2 IV_3 if  industry== `i'
>>>>> }
>>>>>
>>>>> Therefore, there will be 50  time series mean betas, 50 mean R-square
>>>>> for the 50 industries I have. This is easy to execute.
>>>>>
>>>>> The second stage is to take the mean of these 50 time series means
>>>>> beta and the mean of these 50 means R-square, and of course the
>>>>> standard error and the t-stat of mean of the mean betas.
>>>>>
>>>>> In this way, they can obtain the time series mean across industries
>>>>> (rather than firms).
>>>>>
>>>>> Unfortunately, due to my humble knowledge with Stata, I have not
>>>>> figured out the command of how to get stage two (i.e. mean of the mean
>>>>> betas and matched standard error). I tried the following but there is
>>>>> an error
>>>>>
>>>>> forval  i= 1/50 {
>>>>>       statsby _b e(r2)  e(r2_a) , by(industry):xtfmb DV IV1 IV2 IV3 if
>>>>>  industry== `i'
>>>>> }
>>>>>
>>>>> collapse (mean) _b_cons _b_IV1 _b_IV2  _b_IV3  _eq2_stat_1 _eq2_stat_2///
>>>>>  (semean) _se_cons = _b_cons _se_IV1 = _b_IV1  _se_IV2 = _b_IV2
>>>>> _se_IV3 = _b_IV3
>>>>>
>>>>> foreach v in cons IV1 IV2 IV3 {
>>>>> generate _t_`v' = _b_`v' / _se_`v'
>>>>>  }
>>>>>
>>>>> I got error:
>>>>>
>>>>> no; data in memory would be lost
>>>>> r(4);
>>>>>
>>>>>
>>>>> If I remove  by(industry) in the first line of the loop , I got an error as well
>>>>>
>>>>> industry not found
>>>>> r(111);
>>>>>
>>>>>
>>>>> I would highly appreciate it if you can help me to figure out the
>>>>> right command. Also, I do hope this thread to be useful for other
>>>>> people who face this issue as well.
>>>> *
>>>> *   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/