Dear Jorge,
You did not paste the whole output from SPM into your mail, but I
strongly suspect that the output belongs to the comparison between
model 2 and model 1 (in this order!). The numerical value of the
Bayes factor looks fine, and it *does* represent "positive" evidence
(which means 3 < BF < 20 *or* 1/20 < BF < 1/3, depending on the
order of the compared models). You should find the reciprocal of
that Bayes factor and the same statement about positive evidence
further up in the MATLAB window for the the comparison in the reverse
order, i.e. between model 1 and model 2.
Best wishes,
Klaas
At 19:12 24/07/2007, you wrote:
>Dear SPMers,
>
>I am comparing DCM models. All the same VOIs, only difference in intrisic
>connections. But the output from one comparison is odd. I have a BF of
>0.1353 and it states as "positive evidence". I am using the 826 revision of
>SPM. Is this correct?
>
>Many thanks in advance,
>Jorge
>
>All costs are in units of binary bits
>
>All costs are in units of binary bits
>
>Region 1_pos_sp6: relative cost = -2.05e-014, BF= 1
>Region 2_pos_sp6: relative cost = 0, BF= 1
>Region 3_pos_2_sp6: relative cost = 0.0003052, BF= 0.9998
>AIC Penalty = 5.7708, BF = 0.01832
>BIC Penalty = 14.9837, BF = 3.086e-005
>AIC Overall = 5.7711, BF = 0.01831
>BIC Overall = 14.9840, BF = 3.086e-005
>
>Strong evidence in favour of model 1
>Bayes factor >= 0.01831
_____________________________________
Klaas Enno Stephan, MD PhD
Senior Research Fellow, IoN
Wellcome Trust Centre for Neuroimaging
Institute of Neurology (IoN)
12 Queen Square, WC1N 3BG, London, UK
phone: +44-207-8337481
fax: +44-207-8131420
web: http://www.fil.ion.ucl.ac.uk/~kstephan/
|