Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
st: Non-stationarity, first differences and application problems
From
Micha Schildmann <[email protected]>
To
"[email protected]" <[email protected]>
Subject
st: Non-stationarity, first differences and application problems
Date
Sun, 6 Apr 2014 12:43:06 +0100 (BST)
Dear stata-list,
I build a sentiment index and am now validating its statistical properties and significance. I stumbled upon two problems.
(1) -dfgls- indicates that my sentiment is non-stationary as the t-statistics are all above (less negativ) than the critical value. Accordingly, I use the first difference to apply the granger-test. My sentiment now granger-causes the (also non-stationary) market index, which is predicted by theory. But after taking the first difference my sentiment is not able to explain movements in the index using univariate regression, which it did before. Do I have to accept this as it is or do you distinguish between the applicability of first differences in the different tests? For example, would it be possible to use first differences for the granger-causality test and the non-'first differenced' sentiment plus an estimator which takes into account the non-stationarity for the univariate analysis? To my suprise I could not find any literature on non-stationarity and univariate-regressions.
I additionally want to test my sentiment in a panel. Before first differencing my sentiment works really well, afterwards it does not. Here the same question. Can I apply the non-'first differenced' sentiment and use for example -xtpmg- to account for the non-stationarity?
(2) I also checked my dependend variable in the panel data for a unit root using -xtunitroot- and all p-values are 0.000. Therefore, I would judge that all panels are stationary. When I use -dfgls- for seperate panel IDs though, the tests imply non-stationarity for certain lags. Do I misunderstand the test?
Any help is much appreciated!
Best
Micha
*
* 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/