Fixed effect model

[Acorus]

Hi all,

At the moment I am trying to run a fixed effect model for my data, that contains around 10000 companies financial data for years 2005-2009. The problem arises when I specify that cross section effect specification as fixed. The Eviews gives me output, that shows really high R2, but most of my independent variables are statistically insignificant. I get much better numbers (t-test values for independent variables) if I specify period in effects specification as fixed. Could anyone give me any tips what to do in this case (with low t-test values for independent variables) and how to interpret results from both cases?
I would also like to ask, how to add to the regression additional variable (industry), that can get 12 values?
My regression is given below, where ownership and Europe_side serve as dummy variables and help to create 4 subsamples.
total_leverage =c(1)+
c(2)*tangibility+
c(3)*tangibility*ownership+
c(4)*tangibility*europe_side+
c(5)*tangibility*ownership*europe_side+

c(6)*size+
c(7)*size*ownership+
c(8)*size*europe_side+
c(9)*size*ownership*europe_side+

c(10)*profitability+
c(11)*profitability*ownership+
c(12)*profitability*europe_side+
c(13)*profitability*ownership*europe_side+

c(14)*non_debt_tax_shield+
c(15)*non_debt_tax_shield*ownership+
c(16)* non_debt_tax_shield *europe_side+
c(17)* non_debt_tax_shield *ownership*europe_side+

c(18)*age+
c(19)*age*ownership+
c(20)*age*europe_side+
c(21)*age*ownership*europe_side+

c(22)*future_growth+
c(23)*future_growth*ownership+
c(24)*future_growth*europe_side+
c(25)*future_growth*ownership*europe_side+

c(26)*gdp_growth+
c(27)*gdp_growth*ownership+
c(28)*gdp_growth*europe_side+
c(29)*gdp_growth*ownership*europe_side+

c(30)*inflation+
c(31)*inflation*ownership+
c(32)*inflation*europe_side+
c(33)*inflation*ownership*europe_side+

c(34)*corruption+
c(35)*corruption*ownership+
c(36)*corruption*europe_side+
c(37)*corruption*ownership*europe_side
Thanks in advance,

[EViews Glenn]

The reported R2 includes the fixed-effects terms as well (it's not relative to the fixed effects model). There are some forum discussions somewhere around here about computing the latter. Basically, your cross-section effects terms are highly significant.

There's not much one can do about the low t-test values. You could do the F-test to see whether the variables are jointly significant...

I don't understand your last question.