Dear all,
I have one question regarding the tests for correlation among residuals. I have a panel data (7 countries, 15 years). Hausman test indicates I should use Fixed effects model. I've done some tests, including Wooldridge test for autocorrelation in panel data Modified Wald test for groupwise heteroskedasticity, Pesaran's test of cross sectional independence and Breusch-Pagan LM test of independence.
The problem is that Pesaran and Breusch Pagan test give different results. While Pesaran test indicates no correlation among residuals(0.6689), on the other side, Breusch Pagan test says there is correlation among residuals. (0.0012). I am now coonfused as I don't know which one should I use as I have to present the results in my paper? Is there anything I could refer to when choosing one of these two?
Furthermore, if I have the case where there is no correlation between residuals (xtcsd, pesaran abs) but heteroscedasticity(xttest3 )and autocorrelation (xtserial) are present, what should I further do? I've read about method of Driscoll and Kraay, but I don't know how to implement it and what to expect as a result, as I am new to all this.
I hope someone can help on this topic,
Thanks in advance,
Ilda Manjic
I have one question regarding the tests for correlation among residuals. I have a panel data (7 countries, 15 years). Hausman test indicates I should use Fixed effects model. I've done some tests, including Wooldridge test for autocorrelation in panel data Modified Wald test for groupwise heteroskedasticity, Pesaran's test of cross sectional independence and Breusch-Pagan LM test of independence.
The problem is that Pesaran and Breusch Pagan test give different results. While Pesaran test indicates no correlation among residuals(0.6689), on the other side, Breusch Pagan test says there is correlation among residuals. (0.0012). I am now coonfused as I don't know which one should I use as I have to present the results in my paper? Is there anything I could refer to when choosing one of these two?
Furthermore, if I have the case where there is no correlation between residuals (xtcsd, pesaran abs) but heteroscedasticity(xttest3 )and autocorrelation (xtserial) are present, what should I further do? I've read about method of Driscoll and Kraay, but I don't know how to implement it and what to expect as a result, as I am new to all this.
I hope someone can help on this topic,
Thanks in advance,
Ilda Manjic
