Multi-collinearity problem in xtreg ,fe model and process to conduct granger causality test

[Toufiq Sowrov]

I ran -xtreg ,fe- command on my panel data set after running hausman test. FE model omits income variable, which is a binary dummy variable based on country's income level, because of multi-collinearity. Should I drop this variable from my model or keep it? In my linear regression model, fdi estimate was negative which is positive in FE model.

Another thing is I want to test direction of causality among my interest variables. My data are non-stationary. Should I run unit root test and cointegration test first before Granger causality test? I do not know how to conduct this. It would be immensely helpful if I am provided with manual. Thanks

[Carlo Lazzaro]

Toufiq:
1) you seem to deal with a long T (28 years), short N panel (11 panels) dataset. If that were the case, you should switch from -xtreg.fe- to -xtregar,fe-.
2) as expected, the -fe- estimator wipes out time-invarian predictors;
3) you cannot compare the results of -regress- and -xtreg.fe- as they are different beasts and, as such, generate different offspring;
4) see -help xtunitroot- and its many tests.

[nada roua]

Bonjour à tous
J’utilise Stata 16 et j’ai un ensemble de données de panel équilibré avec le temps indiqué par la variable -year - et 11 banques identifiées par le code de la banque.Je veux tester la relation de causalité entre CI , ROA GDP et j’ai eu du mal à le faire .
Existe-t-il une fonction/commande dans Stata qui me permet d’effectuer un test de causalité de Granger sur l’ensemble de données de panel
Merci d’avance à tous.

Salutations sincères

[Carlo Lazzaro]

Nada:
welcome to this forum.
I guess you're looking for -xtunitroot-.