Hello all (again),
Forgive me for barraging you with probably rather silly questions.
However, you haven't told me to shut up yet, so I'll keep going until
you get fed up. :-)
In a recent article in Neuron, Tor Wager et al (2008) states that
"Since our primary concern was group statistics, no autoregressive
(AR) model was used, which are unbiased and valid even without AR modeling."
Since the article gives no reference to support this claim, I was
wondering whether or not it is valid to refrain from using the SPM
defaults with regards to serial correlations.
Best wishes,
Haakon
|