Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | Muhammad Anees <aneesmkhattak@gmail.com> |
To | statalist@hsphsun2.harvard.edu |
Subject | Re: st: RE: RE: RE: Converting unbalanced panel to balanced panel data |
Date | Wed, 16 Mar 2011 14:04:49 +0500 |
Dear All Thank you very much for your suggestion. In real the suggestion of Bryla and Fernando did the job. My assumption was that when I will get a balanced panel, the same individuals will be there in each year. My outcome might be pid year 1 2007 2 2007 3 2007 1 2008 2 2008 3 2008 1 2009 2 2009 3 2009 And indeed the work of Bryla and Fernando did the job. It was only a rough idea if balancing a panel affects the results from the same data which is actually unbalanced. My outcome is somehow insuggestive. So I think balancing is not necessary (may be here in my case only) for the xtnbreg, re and xtnbreg, fe modelling. Now that I have same conclusion for the two types of panel data I might use the unbalanced data. Here I would ask a question for panel data experts in count data. Can I scale the SErrors in xtnbreg output with the deviance or pearson coefficient, so that effects of dispersion is removed? On 16 March 2011 12:10, Jan Bryla <JBR@finansraadet.dk> wrote: > Naturally, you are quite right Dr. Cox. It was an implicit assumption on my part that this would pose no problem for the poster. > > /Jan > > -----Original Message----- > From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Nick Cox > Sent: 15. marts 2011 19:46 > To: 'statalist@hsphsun2.harvard.edu' > Subject: st: RE: RE: Converting unbalanced panel to balanced panel data > > Not quite; this won't automatically produce balance in my understanding of that term. > > You can have the same number of years in two or more incomplete panels but they could be different years. > > Rather, the poster has to say which balance he wants in terms of the trade-off between #panels and #years. > > Nick > n.j.cox@durham.ac.uk > > Jan Bryla > > If you want to drop observations that are not present in all years of your study, generate an indicator for number of observations by pid, such as > > Bys pid: gen nyear=[_N] > Keep if nyear==9 > > Muhammad Anees > > I know I can apply most of the panel data estimations to balanced and > unbalanced panel dataset. Instead of this I would like you all to help > me out of this. > > I have one very basic question in mind and came to no clue after > intensive research of the Stata help files and over the internet. I > have an unbalanced panel dataset and want to make it balanced. I would > present the xtdes results at the end of my email please bear my my > question first: How can I keep if all my id is present at all times so > that I come with atleast balanced panel? > > xtdes > > pid: 10016848, 10016872, ..., 1.794e+08 n = 16368 > year: 2001, 2002, ..., 2009 T = 9 > Delta(year) = 1 unit > Span(year) = 9 periods > (pid*year uniquely identifies each observation) > > Distribution of T_i: min 5% 25% 50% 75% 95% max > 1 1 2 6 9 9 > 9 > > Freq. Percent Cum. Pattern > ---------------------------+----------- > 5110 31.22 31.22 111111111 > 1450 8.86 40.08 ..1111111 > 1187 7.25 47.33 11....... > 616 3.76 51.09 1........ > 486 2.97 54.06 ........1 > 423 2.58 56.65 .11111111 > 410 2.50 59.15 ......111 > 382 2.33 61.49 .......11 > 352 2.15 63.64 ..1...... > 5952 36.36 100.00 (other patterns) > ---------------------------+----------- > 16368 100.00 XXXXXXXXX > > > * > * 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/ > > * > * 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/ > -- Muhammad Anees MSc in Economics The University of Sheffield United Kingdom * * 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/