The Prob > F (<0.0001) of model is much smaller than 0.05 which indicates the most terms of the model including (X1, X2, X3, X4, X2X3, X2X4, X22, X32, X42) are significant (reckoning that the values greater than 0.1 are indication of the model terms are not significant).
The sign of the log likelihood ratio gives an indication of the model that best fits the data (note that its amplitude in absolute value does not have a direct interpretation), but as other statistical quantities, it is sensitive to noise.
This value is generally lower than for the calibration model but gives a better indication of the model's performance.
This value is generally lower than for the calibration but gives a better indication of the model's performance.
Again, this value is generally lower than for the calibration but gives a better indication of the model's performance.
The model output for the non-parametric terms includes an indication of the model "wiggle" expressed as estimated degrees of freedom (EDF).
Although the models are etiologic by nature and not meant as a predictive tool, the percentage predicted correctly are presented as an additional indication of the model fit.
This type of analysis not only gives an indication of the model's predictive strength, but also hints at possible errors in the model (and eventually in the underlying hypotheses from the literature) in the cases where discrepancies between model and observation occur.
While comparison with data is an obvious candidate, an indication of the simulation solving the model equations correctly is if similar results are obtained also from other simulations.
