Dear Pari,

In terms of pooling results from different rats, it is possible that the connectivity (both within source or intrinsic and between source or extrinsic) may be consistent over rats. This is because the phase reversal you describe might be due to the placement of the electrodes. For example, if you have invasive (intralaminar) recordings and you have selected particular contacts, the phase reversal might be due to differential recording depths, in the rats that show phase reversal.

In this context, it would be important to allow for a reversal of the electrode gain (using an LFP forward model with DCM) for each source. The priors allow for different signs; however, you may want to initialise the lead field gain parameter (in DCM.gE.L) to be +1 and -1 for the two sources in the phase reversed animals. This has never been done before but might be a solution, if you are unable to fit the responses accurately with standard priors.

Having inverted each rat separately, you can then use the standard summary statistical approach using the parameter estimates (e.g., t-tests or F tests at the between subject level). Or use recent developments for empirical Bayes and second level (between subjects) analysis:

Empirical Bayes for Group (DCM) Studies: A Reproducibility Study. Litvak V, Garrido M, Zeidman P, Friston K. Front Hum Neurosci. 2015 Dec 22;9:670.

Empirical Bayes for DCM: A Group Inversion Scheme. Friston K, Zeidman P, Litvak V. Front Syst Neurosci. 2015 Nov 27;9:164.

Bayesian model reduction and empirical Bayes for group (DCM) studies. Friston KJ, Litvak V, Oswal A, Razi A, Stephan KE, van Wijk BC, Ziegler G, Zeidman P. Neuroimage. 2015 Nov 11. pii: S1053-8119(15)01037-X

Even if the connectivity parameters differ qualitatively (in sign) between the two groups of rats (with and without phase reversal), you can still perform Bayesian model comparison by pooling the model log evidences. This is because testing for the presence or absence of a connection does not place any constraints on its sign.

I hope that this helps.

With best wishes – Karl

From: pari eghbali [mailto:[log in to unmask]]
Sent: Sunday, January 10, 2016 7:22 PM
To: Friston, Karl
Subject: Phase reversal-DCM

Dear Prof. Friston,

I am going to examine connectivity between 2 signals of some rats’ SEP (between one channel in Vibrissa and one channel in Forelimb areas). In detail, rats stimulated in forelimb peripheral nerve and their evoked potential were recorded from primary somatosensory cortex. In some rats, the two signals are similar in their behavior. Actually they have first pick positive followed by a peak negative while in another rats, the two signal are phase reversal. That is, one of the signals (signal in vibrissa area) against forelimb signal, has first peak negative followed by positive.

I am wondering to know, can I compare rats with each other? Can I put them in same group to decide about general connection between the vibrissa signal and the forelimb signal?

Thank you for your time and consideration,

Regards,

Pari