I have an unbalanced panel dataset with approximately 4000 observations on the return on equity of mfids ranging from year 2003-2010. I am running linear random effects xtregs and read that the within group residuals need to be normally distributed for a RE model to return a useful interpretation. My _roe variable does not meet that requirement.
In order to fix this issue I have tried to add a constant with the size of the minimum value of _roe and then log the resulting observations. Using the xtsktest command my significance value for the "joint test for normality on e" decreases, but still stays within the 5% level.
Is there any remedy to correct for the normality assumption?
Code:
* Example generated by -dataex-. To install: ssc install dataex clear input int year long mfiid float _roe 2007 100000 .2913083 2008 100000 .17713334 2009 100000 .14123334 2004 100001 .04955833 2005 100001 -.008699998 2006 100001 .02328333 2007 100001 .21060835 2009 100001 .015933335 2010 100001 .017058332 2008 100004 .20683333 2006 100008 -.4828166 2007 100008 -.4197917 2008 100008 -2.9300666 2009 100008 9.721633 2005 100012 .3659 2006 100012 -.4431167 2007 100012 1.6969082 2008 100012 1.2119334 2009 100012 .9820334 2010 100012 .16185834 2009 100013 -.10906667 2010 100013 -.13484167 2005 100016 -.0101 2006 100016 .0035833344 2007 100016 .016408335 2008 100016 .05173333 2009 100016 .07153333 2010 100016 .08925833 2003 100017 .17895 2004 100017 .16935833 2006 100017 .09988333 2007 100020 .003608331 2008 100020 .010133334 2009 100020 .09823333 2010 100020 .016058333 2005 100021 -.0768 2006 100021 -.05961667 2007 100021 -.017191667 2008 100021 -.04696666 2009 100021 -.023466665 2010 100021 -.014741667 2006 100024 .9799833 2009 100024 .51533335 2010 100024 .014058333 2004 100026 17.870058 2005 100026 -5.09 2006 100026 9.523084 2007 100026 -2.4865916 2008 100026 .54033333 2009 100026 1.4377333 2010 100026 .4943583 2003 100027 1.0200499 2004 100027 .8740584 2005 100027 .7555 2006 100027 .16448334 2006 100029 -.16821668 2007 100029 .27910832 2008 100029 .07793334 2009 100029 .005633332 2005 100030 -.0131 2006 100030 .024183333 2007 100030 .02750833 2008 100030 .05113334 2005 100031 -.02 2006 100031 -.03771666 2007 100031 -.033691667 2009 100031 .14953333 2010 100031 .19615833 2004 100032 .06425834 2005 100032 -.0216 2008 100032 .17953333 2010 100032 .17735833 2005 100033 .9403999 2006 100033 .7986833 2007 100033 .07110833 2008 100033 .05233334 2009 100033 .03813333 2010 100033 .23085834 2003 100036 1.58285 2004 100036 16.50846 2005 100036 .0053 2006 100036 .6539834 2007 100036 .11130834 2008 100036 -.025066666 2009 100036 -.006966667 2010 100036 .03235833 2003 100039 -.15105 2004 100039 -.10604167 2005 100039 .2303 2006 100039 -.008116666 2005 100041 .6233 2006 100041 .9760834 2007 100041 .25010833 2008 100041 .36893335 2008 100042 -.22946666 2009 100042 -.52516663 2010 100042 -.02404167 2004 100043 -6.986542 2005 100043 .2743 2006 100043 1.2188834 end
Rafael
0 Response to Transform and log return series in a panel dataset to correct for violation of normality assumption
Post a Comment