Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Breusch - Godfrey and Dummy

    Hello everyone,

    I am a Master's student in Banking and Finance. It is my first encounter with Stata and because all professors are overloaded due to the Thesis preparation period I am turning for help here. I have two questions relative to Stata.

    I have a panel data dataset and I have created a dummy variable to check for regionality. I regress with Pooled OLS and then I check for multicollinearity with VIF command and there is no indication for multicollinearity (no variable has a value over 3,5). Then I move to perform a B-G test for autocorrelation by predicting the residuals and finally regressing the residuals against a lag and all my independent variables. The result is an error of 'no observations' and derives from the lagged residuals (l.uhat). The same issue appears with other estimators like fixed effects and random effects.

    The second issue that I have is that when I use the cluster command in all estimators my dummy variable is constantly omitted due to collinearity. Let me repeat that even in pooled OLS where I perform a vif check there is no problem but the cluster command omits the dummy every time. I tried to do the i.dummy command but I saw no change whatsoever.

    I request your help. My panel data are for 10 Countries and for each country time is 520 observations (weekly observations for 10 years).

    Thank you in advance,
    Giorgos




  • #2
    Giorgos:
    welcome to this forum.
    With no further details/data excerpt (see CODE delimiters and -dataex- in FAQ) from your side, the only thing that springs to my mind is that you can rely on better estimator than pooled OLS (if there's evidence of a panel-wise effect) for long panel dataset, such as -xtregar- and -xtgls-.
    Kind regards,
    Carlo
    (Stata 19.0)

    Comment

    Working...
    X