# System Identification Toolbox - How is the quality of the model output calculated?

I'm searching for the calculation of the quality of the estimated model output compared to the measured data. I'm working with the System Identification Toolbox. You can see an example here http://de.mathworks.com/help/ident/ref/compare.html at the first graph. Is the percentage calculated with an integral or any other method?

I found the answer. It is at the same page at Output Arguments --> Fit. The fit is calculated (in percentage) using: $$fit=100*(1-\frac{||y-\hat{y}||}{||y-mean(y)||})$$ where $\hat{y}$ is the estimated model and $y$ the measured data