Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | Meredith Herd <meredithherd@yahoo.com> |
To | "statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu> |
Subject | st: Using sureg or xtsur for a SUR with unbalanced panel |
Date | Thu, 27 Feb 2014 04:09:15 -0800 (PST) |
This is my first-time posting to statalist and I am developing my statistical skills, so many thank you's for considering my questions. I have an unbalanced panel (52 countries, 335 observations, unbalanced over 2002-2011) – currently arranged in long form. I have a system of 5 equations where the regressors are identical, except for a lagged dependent variable. (y1 x1 x2 x3 x4 L1.y1) (y2 x1 x2 x3 x4 L1.y2) …(y5 x1 x2 x3 x4 L1.y5) Because my panel is unbalanced, I will use the user-written command -xtsur- which, if I understand correctly, should be organised in long form per http://fmwww.bc.edu/repec/bocode/x/xtsur.html(i.e. “N_1 individuals observed once come first, N_2 individuals observed twice come second, etc. such that the set of N_1 individuals is considered as a cross section, and sets of N_p individuals where p = 2,...,P as balanced panels”). This is in contrast to -sureg- command which I read must have data arranged in wide form. I have done this using the following commands: by panelid: gen CountN=_N by panelid: gen Countn=_n gsort countN panelid year drop countN tsset panelid countn Q1: Is it correct if using -xtsur- that my time variable should be countn as opposed to year when tsseting? Q2: What Stata 13.0 syntax should I use to constrain my regression such that my 5 dependent variables sum to 1? From the help files, I know it is - constraint define 1 - but I am not sure how to define this/what syntax should follow this. Q3: When setting my data as a time series, should I use tsset or xtset? Is there any difference? Q4: xtsur does not calculate adjusted R-squared, are there commands I could run to find this? Q5: Are there diagnostic commands I can perform post -xtsur- to check for autocorrelation, multicollinearity (e.g. vif), heteroscedasticity? * * 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/