i am using panel data with three groups A=15 B =37 C =12 cross section data 1990-2019 running the following codes in stata 14
xtpmg d.lip d.lop d.lex, ec(ec) lr(l.lip lop lex ) mg replace
xtpmg d.lip d.lop d.lex, ec(ec) lr(l.lip lop lex ) pmg replace
the first one for mean group latter for pooled mean group for group A and C i get results from stata for mg and pmg for B group mg results appears whereas pmg results keep on iterating with ml log likelihhod iteration goes to 3000 but no results for day or two till i break it .
i am unable to understand reason does any one have idea about .
will be highly obliged
This is where in my opinion, eviews is superior to stata. Try running on eviews and see the ease.
I can tell you 2 reasons for not using the xtpmg package.
First one, I know that you attempted to use the mean group ARDL, pool mean group ARDL and DFE-ARDL by Pesaran and Shin (1995), but the package use the Maximum likelihood estimate (MLE) so it might cost you pretty much time to process. Besides, the PMG, MG and DFE do not account for the problem of cross-sectional dependence. As far as I know, the package xtdcce2 (Dynamic common correlated effects) by Jan Ditzen (2018) is way more efficient in dealing with both cross-sectional dependence and slope heterogeneity. One more advantage of xtdcce2 is that it does not use the MLE so you will not be worry 'bout frustratedly sitting in front of the screen waiting for the results
Second, the residuals returned from xtpmg are not centered around zero, as they are theoretically supposed to. Sometimes, you have to extract the residual from the model for further investigation such as testing for the presence of cross-sectional dependency. My suggestion is using the xtmg AMG or CCEMG instead of xtpmg in case if you wanted to address the slope heterogeneity.
- Badges
- Science topic
- Similar topics
- Biological Science
- Histology
- Sectioning