Thank you for your informative videos. I have a question about choosing between the ARDL and FGLS models given my data's characteristics (T>N, but T=16 and N=7, heteroskedasticity, and slope heterogeneity). In this case, can the FGLS model be used in place of the ARDL model when some variables are stationary at I(1) and I(0) and there is cointegration between them? Sorry if my question sounds basic. I am new to these types of models. Appreciate your insights.
Good Evening Sir. Thank you so very much for your videos - they are always so well explained. I especially like the fact that you tell us of the need to install the programs needed so we can execute the estimation commands. Please with regards to the estimation for mg (mean group) - when I execute the command, it returns with the following error message: "invalid new variable name. Variable name ec is in the list of predictors." Please how can this be resolved? I look forward to your kind revert. Thank you.
Hi Angie. I hope you day is going wey. It's a common problem with Stata 16 when you're MG. Still trying to find a way around the version. However, it works well with Stata 14 and 15. You might want to consider them. Regards
@@josephlanre Thank you so very much Sir for your kind reply. If I may ask, if one or two of the variables are not stationary at level (for the IPS unitroot test) but are stationary at first difference - do you still go ahead to estimate or do you have to denote the first difference of the variable in your long run variables. For example, going by your model, if your GDPG and INF are stationary at first difference do you go ahead to estimate it like you've shown us or do you estimate it like this: xtpmg d.dgdpg d.fdi d.dinf d.exchr, ec(ec) lr(l.dgdpg fdi d.inf exchr) replace Thank you so very much Sir.
@@josephlanre thank you very much. I am trying to use the ccepmg as i have cross sectional dependence in my variables. Since the xtcce2 command in stata displays both short run and long sun results. If i got you correct i should only report the short run section and regard the long run section as insignificant right? all variables from the long run section are insignificant. Apologies for the long comment.
Thank you for your informative videos. I have a question about choosing between the ARDL and FGLS models given my data's characteristics (T>N, but T=16 and N=7, heteroskedasticity, and slope heterogeneity).
In this case, can the FGLS model be used in place of the ARDL model when some variables are stationary at I(1) and I(0) and there is cointegration between them?
Sorry if my question sounds basic. I am new to these types of models.
Appreciate your insights.
Yes, That would be correct. FGLS model is potent, especially with variables exhibiting heteroskedasticity and or autocorrelation
Good Evening Sir. Thank you so very much for your videos - they are always so well explained. I especially like the fact that you tell us of the need to install the programs needed so we can execute the estimation commands. Please with regards to the estimation for mg (mean group) - when I execute the command, it returns with the following error message: "invalid new variable name. Variable name ec is in the list of predictors." Please how can this be resolved? I look forward to your kind revert. Thank you.
Hi Angie. I hope you day is going wey. It's a common problem with Stata 16 when you're MG. Still trying to find a way around the version. However, it works well with Stata 14 and 15. You might want to consider them. Regards
@@josephlanre Thank you so very much Sir for your kind reply. If I may ask, if one or two of the variables are not stationary at level (for the IPS unitroot test) but are stationary at first difference - do you still go ahead to estimate or do you have to denote the first difference of the variable in your long run variables. For example, going by your model, if your GDPG and INF are stationary at first difference do you go ahead to estimate it like you've shown us or do you estimate it like this: xtpmg d.dgdpg d.fdi d.dinf d.exchr, ec(ec) lr(l.dgdpg fdi d.inf exchr) replace
Thank you so very much Sir.
Hello sir, pmg and MG both have long run and short run, what is how we interpret these results of pmg and MG results of long run and short run
great video.how do i proceed with panel heterogeneous when there is no cointegration
No cointegration means no long-run relationship. You'll have to focus on the short-run component of your result, and not interpret the long run
@@josephlanre which model would be best?i have mixed variables of I(0) and I(1)
@@lombechangala2803 ARDL model
@@josephlanre thank you very much. I am trying to use the ccepmg as i have cross sectional dependence in my variables. Since the xtcce2 command in stata displays both short run and long sun results. If i got you correct i should only report the short run section and regard the long run section as insignificant right? all variables from the long run section are insignificant. Apologies for the long comment.
Good day Sir, please when I put in the mg command it says ‘invalid new variable ec in the list of predictors’ what can I do to stop this error
Hi Williams. I assume you are using Stata 16. It's a common problem with that version. Try it with Stata 14 or 15.
Regards