Generalizability vs. precision
•
For each extra parameter we include in a model, we
improve the fit:
But the generalizability suffers if we fit to noise
) ( max ),
( max
,
XL
bX
L
X
b