pooled regression for panel data
Posted: Sat Oct 13, 2012 10:59 am
hi,
i'm new in eviews and i hope somebody can help me..
i have 248 firms observations each year for the period of 2 years (2006 and 2009). thus total sample is 496 firm-years observations.
1 dependent variable, 9 independent variables(IV) and control variables(CV).
the IV and CV consist of continuous and dummy variables.
my problem is, one of my dummy variable have constant data for both year 2006 and 2009. example as follows;
VARIABLE: FINANCIAL YEAR END
COMPANY A (2006) score=1
COMPANY A (2009) score=1
when i run Hausman test, the result shows significant -> means that i supposed to use FIXED EFFECTS.
however, due to my FINANCIAL YEAR END dummy variable (in which the score constant for both year 2006 and 2009), eviews pop up the notification of "near singular matrix"
my question is;
1. if i did not select either 'random effects' or fixed effects (i just leave it as "none"), is that considered as pooled regression in panel data?
2. is it the right way if i leave the "effects specification" as "none" and then i select "white cross-section" in "coef covariance matrix" in order to solve the heterokedasticity problem in my data?
kindly advice...thank you.
i'm new in eviews and i hope somebody can help me..
i have 248 firms observations each year for the period of 2 years (2006 and 2009). thus total sample is 496 firm-years observations.
1 dependent variable, 9 independent variables(IV) and control variables(CV).
the IV and CV consist of continuous and dummy variables.
my problem is, one of my dummy variable have constant data for both year 2006 and 2009. example as follows;
VARIABLE: FINANCIAL YEAR END
COMPANY A (2006) score=1
COMPANY A (2009) score=1
when i run Hausman test, the result shows significant -> means that i supposed to use FIXED EFFECTS.
however, due to my FINANCIAL YEAR END dummy variable (in which the score constant for both year 2006 and 2009), eviews pop up the notification of "near singular matrix"
my question is;
1. if i did not select either 'random effects' or fixed effects (i just leave it as "none"), is that considered as pooled regression in panel data?
2. is it the right way if i leave the "effects specification" as "none" and then i select "white cross-section" in "coef covariance matrix" in order to solve the heterokedasticity problem in my data?
kindly advice...thank you.