cross sectional regression wls
Posted: Wed May 19, 2010 1:32 am
Hi!
I am doing a cross sectional analysis on credit spreads from different firms. I have to run a dummy variable regression on countries and industries (so two sets of different dummy variables), also including some control variables (leverage, firm size etc.).
Credit spread = α + Σβ*I + Σγ*C + Σδ*F + ε
where β accounts for the industry factor in the dummy variable regression and γ accounts for the country factor.
In order to avoid perfect multicollinearity (dummy variable trap), I have been told to include the following restrictions:
Σw*β = 0
Σv*γ = 0
, where w and v are weights previously calculated (the weight of the particular country/industry in the market portfolio).
I think that I have to use the weighted least squares method, but I am not sure. Moreover, I do not know how I can impose these 2 restrictions simultaneously.
Can anybody help me? I really need to find out how to do this.
Thanks!
I am doing a cross sectional analysis on credit spreads from different firms. I have to run a dummy variable regression on countries and industries (so two sets of different dummy variables), also including some control variables (leverage, firm size etc.).
Credit spread = α + Σβ*I + Σγ*C + Σδ*F + ε
where β accounts for the industry factor in the dummy variable regression and γ accounts for the country factor.
In order to avoid perfect multicollinearity (dummy variable trap), I have been told to include the following restrictions:
Σw*β = 0
Σv*γ = 0
, where w and v are weights previously calculated (the weight of the particular country/industry in the market portfolio).
I think that I have to use the weighted least squares method, but I am not sure. Moreover, I do not know how I can impose these 2 restrictions simultaneously.
Can anybody help me? I really need to find out how to do this.
Thanks!