1066-1067 / 1676
Data-driven methods
•
based on improvement of some measure of
goodness of fit / strength of relation to data
when adding a parameter:
–
Coefficients p-value in multivariate analysis
–
Improvement of likelihood in max. likelihood fit
–
…
•
Advantage: Flexible
–
Any candidate predictor can be considered
•
Main challenge: Parameter selection!