Dear SPM experts,
We are evaluating two possible models at the first level, one with 13
regressors and the other with 17 regressors (4 regressors more).
Our aim is to evaluate if there is a significant F change when we add
the 4 additional regressors.
The formula to calculate the F change would be:
[(EF-ER)/(dfF-dfR)]/(EF/dfF)
where EF is the Residual Error of the full model (17 regressors) and
would correspond to SS(1-R^2 of EF)
ER=Residual Error of the model with 13 regressors = SS(1-R^2 of ER).
df=degrees of freedom
Our questions:
1. Is the above formula correct?
2. How can we get the value of Residual Error in SPM? Are we supposed to
use the ResMS image?
3. Is this something that could be specified in a anova Ftest analysis
in SPM second level? If we look at the F score of significant clusters
at the second level and we compare the F scores of the same clusters in
the 2 models (subtracting them), do we already have a value of the F
change?
Any suggestion much appreciated!
Thanks
Daniela
|