¥F-test for Lack of Fit (lof)
HIf multiple measurements are
available (and they should be), a Lack Of Fit (lof) test is first
carried out.
åHypothesis:
H
H0: E(Y) = b0 + b1X1 +
b2X2 +
É,+ bp-1Xp-1
H
Ha: E(Y) b0 + b1X1 +
b2X2 +
É,+ bp-1Xp-1
å
åHypothesis is tested by
comparing the variance of the modelÕs lack of fit to the measurement
variance at each point (pure error).
åIf Flof is significant then model is
inadequate. STOP HERE.
íReconsider independent
variables, try again.
åIf Flof is insignificant then model appears
adequate, so far.
åIt is important to test for
the lack of fit:
íThe remainder of the
analysis assumes an adequate model is used
íYou will not be visually
inspecting the goodness of the fit for thousands of voxels!