st: RE: Build groups with the same first two numbers of SIC

["Nick Cox" <n.j.cox@durham.ac.uk>]

I can't see any reason whatsoever here for your strategy of separating into smaller datasets and then putting them together. 

I'll guess that -sic- is numeric. 

You need a variable, say -sicgroup-, that you put together the hard way 

gen sicgroup = 2 
replace sicgroup = 1 if sic == 3674 
replace sicgroup = 3 if sic == 3861 
replace sicgroup = 4 if sic == 4213 

Then it's just 

egen mean = mean(ret), by(sicgroup date) 

Nick 
n.j.cox@durham.ac.uk 

Hua Pan

I have a list of firms with four digit sic code, permno (identify Nr. for firms), date and return and wish to get daily mean return within the group, which has the same first two numbers of SIC code.

My Dataset look like this:

sic     permno      date          ret 
…
3674   10012    5.Jan.2004
        10012    6.Jan.2004
       10012    7.Jan.2004
3674   10259    5.Jan.2004
       10259    6.Jan.2004
       10259    7.Jan.2004
3674   10299    
       10299    
       10299  
3674   10302
       10302
       10302
-----------------------------------------------------------------
3714   10667
       10667
       10667
------------------------------------------------------------------
3728   10145
       10145
       10145
------------------------------------------------------------------
3861   10163
       10163
       10163
------------------------------------------------------------------
4213   10379
       10379
       10379
4213   10649
       10649
       10649


At first I want to build several groups. Firms within each group have the same character: the first two numbers of their SIC codes are identical. For the example above   
        sic               permno
Group1: 3674              10012, 10259, 10299, 10302
Group2: 3714, 3728        10667, 10145
Group3: 3861              10163
Group4: 4213              10379, 10649

Then I wish to get mean daily return for each group.

So I just tried to separate the big dataset into several sub dataset, and calculate daily mean return for each of them. Then I get the sub datasets together with “append”. For the first step, I did: 

. local n=3600
. while `n' <4300 {
  2. use "D:\sic.dta", clear
  3. keep if sic >=`n' & sic < `n'+100
  4. by date, sort: egen meanret=mean(ret) 
  5. save "D:\ph\sic\sic_`n'.dta"
  6. local n=`n'+100
  7. }


It is successful for 36xx, 37xx. But when `n’== 3900, all observations in the complete file “D:\sic.dta" are deleted, because none of them meet the requirement: sic>=3900 & sic < 4000, so there is an error: 

(y observations deleted)
__000001 not found

y is the number of all the observations in complete dataset.

There are a huge number of observations, so I can’t do it one by one. Has anyone here an idea to solve this problem? Or some easier methods to generate such groups (I’ve also tried, but failed to get it ), so I can get the daily mean return with:

by group date, sort: egen meanret=mean(ret)

Btw, I’m using stata 10.

*
*   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/