Dear SPM crew,
I'd like to draw meaningful error bars around plots of DCM parameters (A and B matrices), computed using fixed-effects Bayesian model averaging. The plots represent connection strengths for two conditions - baseline [ plotted value calculated as exp(bma.Ep.A) ] and a modulated condition [ exp(bma.mEp.A+bma.mEp.B) ] . I'd be very grateful if you could have a look at the following questions:
(1) the group standard deviation values (bma.mEps) differ from standard deviations of single subjects' mean parameter values [ std(bma.sEp) ] . I expect I should use the former in plotting error bars, am I correct?
(2) In DCM output, trial-specific effects are plotted as exponentials of parameter values in A and B matrices. However, calculating exponentials of standard deviations alone would make their values disproportionately large, relative to the mean parameters. Do you think it's fine if, for the baseline condition, I calculate exp(bma.mEp.A ± bma.sEp.A] instead? How about the modulated condition - how do I integrate std values from A and B matrices?
(3) To possibly plot SEMs instead of standard deviations, is simply dividing std values by square root of sample size OK in the fixed-effects Bayesian model averaging context?
Thanks a lot in advance!
Ryszard Auksztulewicz
Berlin School of Mind and Brain
Humboldt-Universität zu Berlin
|