Is Regression in Excel wrong?

Hello,

I have tested Simple Linear Regression on the following dataset that made Excel 10 produce nonsense results. I am also beta testing Excel 11, and there the results seem OK, however, the adjusted R2 is different in Excel 11 compared to statistXL! I have not manually checked the results (the constant is set to zero in this analysis). Is Excel 11 wrong or is StatisiXL! wrong (or both)?

Dataset
X Y
2008 11300
1909 12400
1351 5500
1892 60500
4312 12700
2548 6500
2236 87800
1803 26000
2088 4300
743 69500
3291 7200
2079 44000
2468 131000
1187 8000
2932 7900
1933 28600

Regards,

Rasmus

Comments

  • Hi Rasmus

    I have checked out your dataset and can assure you that the R2 given by statistiXL is correct (0.354). Excel 10 and prior all seem to give wrong results for this analysis. The Excel 11 beta is better but not perfect. It actually gives the correct R2 but the incorrect Adjusted R2 and Probability value for F in the ANOVA results. It looks as though they are not adjusting their degrees of freedom when the constant is forced through the origin.

    Cheers

    Alan
Sign In or Register to comment.