Dear Lexi, Thanks a lot for your prompt answers, The issue i'm confronted to is the following: i have a panel data N=17 T=5 (annual observations) and wanted to check for stationarity to avoid a spurious regression, but the question is do i' have the right do do so??? it's statistically correct? if no is there any alternative method to verify if our regression is correct?
Thanks again Ama == Dear Ama, I copy my reply to the list, in case someone needs it. Spurious regression occurs when correlation between time series variables results from their common trends - the variables tend to move together over some cycle. However, it may difficult to decipher whether or not the variables in your model have significant trends; also trends differ (see Enders 1995 Time Series Econometrics). So to deal with this, you must perform formal integration tests. If the variables have unit root (ie, non-stationary) then you cannot model the variables in their levels. You must transform them by appropriate differencing. Then you can model using a dynamic model or error-correction model (ecm) (if the variables are cointegrated). Use of ecm makes sense only if the time span of your data is "long enough" - it is a long run concept. Long enough depends on the phenomenon under study. If theory suggests that equilibrium could occur within the time span of your data (17 years in your case - this is long enough in most cases), then concepts of cointegration and ecm are relevant. Hope this helps. Lexi -----Original Message----- From: r-help-boun...@r-project.org [mailto:r-help-boun...@r-project.org] On Behalf Of Setlhare Lekgatlhamang Sent: Saturday, July 24, 2010 1:01 PM To: amatoallah ouchen; r-help@r-project.org Subject: Re: [R] Question regarding panel data diagnostic Let me correct an omission in my response below. The last sentence should read "But if the data are 10 quarterly or monthly values, these techniques are not relevant". Cheers Lexi -----Original Message----- From: r-help-boun...@r-project.org [mailto:r-help-boun...@r-project.org] On Behalf Of Setlhare Lekgatlhamang Sent: Saturday, July 24, 2010 12:54 PM To: amatoallah ouchen; r-help@r-project.org Subject: Re: [R] Question regarding panel data diagnostic My thought is this: It depends on what you have in the panel. Are your data cross-section data observed over ten years for, say, 3 countries (or regions within the same country)? If so, yes you can perform integration properties (what people usually call unit root test) and then test for cointegration. But if the data are quarterly or monthly, these techniques are not relevant. Hope this helps. Lexi -----Original Message----- From: r-help-boun...@r-project.org [mailto:r-help-boun...@r-project.org] On Behalf Of amatoallah ouchen Sent: Friday, July 23, 2010 12:18 AM To: r-help@r-project.org Subject: [R] Question regarding panel data diagnostic Good day R-listers, I'm currently working on a panel data analysis (N=17, T=5), in order to check for the spurious regression problem, i have to test for stationarity but i've read somewhere that i needn't to test for it as my T<10 , what do you think? if yes is there any other test i have to perform in such case (a kind of cointegration test for small T?) Any hint would be highly appreciated. Ama. * * 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/ ______________________________________________ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. DISCLAIMER:\ Sample Disclaimer added in a VBScript.\\ .....{{dropped:15}} ______________________________________________ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
