Different result Eviews vs Excel

[quoctruongphan]

Hello,

I am running a panel regression, the R^2 between Eviews and Excel is kind of different.
R^2 Eviews: -0.033164
R^2 Excel: 0.184428

To my knowledge, SST=SSR+SSE and all these value are positive. Two questions here is:
1. Why R^2 in Eviews could be negative?
2. Why there is such different between Excel Regression and Eviews in calculating R^2 while all coefficients between two programs are identical?

I included:
- Eviews file. Just take a look at eq01.
- Excel file.
+ Sheet1: all data
+ Sheet2: sample used for estimate Regression the same as sample used in Eviews. Data filter: GISC=4 and doup=1

Thank you for viewing

===============================
Unrelated questions:
1. Can we run fixed effect regression without constant c?
2. In Estimate Equation, how can we filter data >90th percentile and data<10th percentile

[quoctruongphan]

Hope someone can help this topic,
Many thanks

[EViews Gareth]

1. You have no constant. R^2 can go negative if you have no constant.
2. Without knowing how Excel calculates R^2 in the no-constant model, it is hard to know how they are differing.
3. You can specify an equation with fixed effects without a constant, but EViews will add it in for you.
4. Use the sample

[quoctruongphan]

1. You have no constant. R^2 can go negative if you have no constant.
2. Without knowing how Excel calculates R^2 in the no-constant model, it is hard to know how they are differing.
3. You can specify an equation with fixed effects without a constant, but EViews will add it in for you.
4. Use the sample

Thank you for your answer but:
SSE between Eviews & Excel have the same estimation: 5.06E+08 but different SSR:
+ Eviews: SSR=2383.93
+ Excel: SSR=1.14E+08

and general formula to calculate R^2 is: SSR/SST = SSR/(SSR+SSE)
Here, no matter how substituting, there is no way to obtain a negative R^2 here

Eviews: 2383/(2383+5.06E+08) ~ 3.84E-06 (non negative)
Exel : 1.14/(1.14+5.06) = 0.18

I don't know if there is any mistake in my statement, if not then can you explain more how could without constant, R^2 can be negative?

Thank you for your answer

[EViews Gareth]

That formula doesn't apply. To see how EViews calculates R-squared, see User Guide II.