2

I am working on a study of fire propagation risk, and I have estimated a generalized linear model. Looking at the resulting ANOVA, I have seen that the proportion between the model sum of squares and the total sum of squares does not match the $R$-squared terms displayed in the model output. The ANOVA is:

enter image description here If I have calculated it correctly, the $R$-squared term should be:

$R^{2} = \frac{\text{Model SS}}{\text{Total SS}} = \frac{26.64}{40.93} = 0.65$

I am also unsure about what the result should be expected (0.75 as in the multiple $R^{2}$, or 0.68)?

Do you know why this sum does not match with the $R$-squared that I computed from the table? Possibly am I confusing concepts?

Alexis
  • 26,219
  • 5
  • 78
  • 131
user33045
  • 21
  • 1
  • too long didn't read. you can check if it is a intercept problem, see here https://stats.stackexchange.com/questions/233256/why-i-am-getting-different-r2-from-r-lm-and-manual-calculation/233257#233257 – Haitao Du Jan 24 '18 at 19:29
  • R2 = 1.0 - (absolute_error_variance / dependent_data_variance) should yield the correct value. – James Phillips Jan 24 '18 at 19:44
  • Thanks very much for your comments @hxd1011 and James Phillips I have read about the intercept problem and I think you are right, so probably the solution would be to compute the R2 as James Philips suggests. However, I have an important problem, that I have lost the original database, so... is there any form to compute the variances from these sums of squares in the table? Thanks very much. – user33045 Jan 25 '18 at 14:19

0 Answers0