Panel data

[Momoko Yamamoto]

I hope all is well and stay safe at home.

I am a very beginner for STATA. I currently clean administrative firm-level data. I would like to create panel data for a certain period of years.

Each firm reported its personal information in several fiscal years. They have individual “ID” and we can see “FY”, which is the fiscal year when they declared the information.

For instance, some of these firms started reporting their income from 2006 till 2018 each year but some firms did not report it each year, others reported in a few years.
It means there are different sets of panel data.

Hence, I would like now to know which period of years has a large number of observations so that I can use a certain period of years to do further analysis. I need to include the information of FY 2018 (the latest) So, I would like to know how many numbers of observations for each combination of years of panel data.
In total, there are currently around 100,000 observations. below is an extract.

Would you mind sharing your knowledge on how I can figure it out...? I sincerely appreciate your support.


ID Sector FY
1733 Servicios 2006
1733 Servicios 2007
1733 Servicios 2008
1733 Servicios 2009
1733 Servicios 2010
1733 Servicios 2011
1808 Agropecuaria 2006
1808 Agropecuaria 2007
1808 Agropecuaria 2008
1808 Agropecuaria 2009
1808 Agropecuaria 2010
1808 Agropecuaria 2011
1808 Agropecuaria 2012
1808 Agropecuaria 2013
1808 Agropecuaria 2014
1808 Agropecuaria 2015
1808 Agropecuaria 2016
1808 Agropecuaria 2017
1808 Agropecuaria 2018

2541 Servicios 2014
2541 Servicios 2015
2541 Servicios 2016
2541 Servicios 2017
2541 Servicios 2018

3301 Servicios 2018

3480 Industrias 2007
3480 Industrias 2008
3480 Industrias 2009
3480 Industrias 2010
3480 Industrias 2011
3480 Industrias 2012
3480 Industrias 2013
3480 Industrias 2014
3480 Industrias 2015
3480 Industrias 2016
3480 Industrias 2017
3480 Industrias 2018

3594 Servicios 2014
3594 Servicios 2015
3594 Servicios 2016
3594 Servicios 2017
3594 Servicios 2018

4854 Servicios 2006

5401 Servicios 2015
5401 Servicios 2016
5401 Servicios 2017
5401 Servicios 2018

6586 Servicios 2006
6586 Servicios 2007
6586 Servicios 2008
6586 Servicios 2009
6586 Servicios 2010
6586 Servicios 2011
6586 Servicios 2012
6586 Servicios 2013
6586 Servicios 2014
6586 Servicios 2015
6586 Servicios 2016
6586 Servicios 2017
6586 Servicios 2018

6885 Servicios 2012
6885 Servicios 2013
6885 Servicios 2014
6885 Servicios 2015
6885 Servicios 2016
6885 Servicios 2017
6885 Servicios 2018

7385 Servicios 2006
7385 Servicios 2007
7385 Servicios 2008
7385 Servicios 2010

7496 Industrias 2016
7496 Industrias 2017

7839 Servicios 2010

7951 Servicios 2006
7951 Servicios 2007
7951 Servicios 2008
7951 Servicios 2009
7951 Servicios 2010
7951 Servicios 2011
7951 Servicios 2012
7951 Servicios 2013
7951 Servicios 2014
7951 Servicios 2015
7951 Servicios 2016
7951 Servicios 2017
7951 Servicios 2018

8063 Industrias 2006
8063 Industrias 2007
8063 Industrias 2008
8063 Industrias 2009
8063 Industrias 2010
8063 Industrias 2011
8063 Industrias 2012
8063 Industrias 2013

8333 Servicios 2006
8333 Servicios 2007
8333 Servicios 2008
8333 Servicios 2009
8333 Servicios 2010
8333 Servicios 2011
8333 Servicios 2012
8333 Servicios 2013
8333 Servicios 2014
8333 Servicios 2015
8333 Servicios 2016
8333 Servicios 2017
8333 Servicios 2018

8923 Servicios 2011
8923 Servicios 2012
8923 Servicios 2013
8923 Servicios 2014
8923 Servicios 2015

[Andrew Musau]

What you have is an unbalanced panel and most panel estimators handle unbalanced panels. Therefore, your reason for getting rid of observations is not sound and may ultimately bias your analysis. You should make use of all observations whenever you can. In any case, here is a way of finding out the most frequent time range.

bys ID (FY): gen range= string(FY[1])+"_"+string(FY[_N])
tab range, sort