 |
 |
To prevent any misinterpretation, let me extend the 2nd sentence from my previous email below: "However, this does not prevent you from making inferences on parameter estimates across subjects, provided these estimates are based on the same model in each subject." Best wishes Klaas ----- Weitergeleitete Mail ---- Von: Klaas Enno Stephan <[log in to unmask]> An: [log in to unmask] Gesendet: Mittwoch, den 13. Oktober 2010, 18:02:58 Uhr Betreff: Re: [SPM] DMC Question - Please help!!! Dear Branislava You are absolutely right that any (function of) model parameter estimates is conditional on the model chosen; this concerns both absolute (additive) and relative (ratio) interpretations of how B relates to A. However, this does not prevent you from making inferences on parameter estimates across subjects. This inference is usually on (linear contrasts of the) estimates within parameter classes (i.e., A, B C or D). You might also be able to motivate a second-level frequentist test on across-class mixtures such as m=(A+uB) because, with zero-mean priors as in DCM, m will be zero under the null hypothesis of no effect across subjects. However, this is not common practice and requires that (the average of) u is identical across subjects. In short: There is nothing wrong with either the absolute or relative interpretation of how B relates to A. However, both interpretations are just used to help readers understanding the results, they do not usually form the basis for inference. I hope this clarifies things. All the best Klaas ________________________________ Von: "Curcic, B (med)" <[log in to unmask]> An: Klaas Enno Stephan <[log in to unmask]> Gesendet: Montag, den 11. Oktober 2010, 15:45:13 Uhr Betreff: RE: [SPM] DMC Question - Please help!!! Dear Klaas, Thank you for your answer. I understand that both (0.2 and 60%) are equivalent in a mathematical way (and I understand the unity restriction for inputs). However, phenomenologically, the absolute values of coupling parameters (such as 0.5 or 1.38 etc) standing alone are not adding to our information because those values are restricted to the particular DCM model. As far as I understand only relative values (to those within the same model) are of importance. For example, if I look at the same areas, and coupling parameters for the same connections between models, they are different (say 0.72 in DCM1 and 0.52 in DCM2). Those coupling parameters vary more between models than between subjects within the same model. Therefore, I am not sure that summing them add to overall information. However looking at the relative value such as 60% has a phenomenological meaning. Or do I miss something in my reasoning? Thank you very much Kind regards Branislava -- Branislava Curcic-Blake, PhD Cognitive Neuropsychiatry group BCN Neuroimaging Center (NIC) University Medical Center Groningen Antonius Deusinglaan 2 9713 AW Groningen Tel: +31503638806 Fax: +31503638875 [log in to unmask] -----Original Message----- From: Klaas Enno Stephan [mailto:[log in to unmask]] Sent: Mon 11-10-2010 15:14 To: Curcic, B (med); [log in to unmask] Subject: RE: [SPM] DMC Question - Please help!!! Dear Branislava, These interpretations are equivalent. You could either say that the presence of the modulatory input (assuming it has an amplitude of unity) reduces the effective connectivity from 0.5 to 0.2 Hz, or that it leads to a reduction of effective connectivity by 60%. Note the dependence on the amplitude of the modulatory input. This becomes particularly important when dealing with parametric regressors whose amplitudes change over time (e.g., models of trial-by-trial learning). Best wishes Klaas ________________________________ Von: "<Branislava> <Curcic-Blake>" <[log in to unmask]> An: [log in to unmask] Gesendet: Montag, den 11. Oktober 2010, 11:00:03 Uhr Betreff: Re: [SPM] DMC Question - Please help!!! Hello Fred and Fu, I am not sure that A's and B's should be added like that. In principle they have the same dimensions, (1/second) but the meaning of summing or subtracting them is not clear to me specially when their absolute values are as far as I understand not important. As far as I understand their relative value is meaningful, so in your case Fred, you would be interested in their ratio (-0.3)/(0.5)= -0.6. That can be interpreted as 60 % decrease of the effective connectivity due to the modulatory effect. Kind regards Branislava De inhoud van dit bericht is vertrouwelijk en alleen bestemd voor de geadresseerde(n). Anderen dan de geadresseerde(n) mogen geen gebruik maken van dit bericht, het niet openbaar maken of op enige wijze verspreiden of vermenigvuldigen. Het UMCG kan niet aansprakelijk gesteld worden voor een incomplete aankomst of vertraging van dit verzonden bericht. The contents of this message are confidential and only intended for the eyes of the addressee(s). Others than the addressee(s) are not allowed to use this message, to make it public or to distribute or multiply this message in any way. The UMCG cannot be held responsible for incomplete reception or delay of this transferred message.