Dear Pari,
As we have discussed before, you can't really define a model in which
the connection strength only changes in one direction (in theory you
could do that, just it's not supported in DCM code) So I suspect your
two models are actually the same model just with inverted sign of the
B parameter. So I suggest that you report that you only had one model
and used two-sided t-test to test for direction of the effect.
You should do the test on Ep. Pp is for Bayesian inference but if you
are doing a classical test it's not relevant.
Best,
Vladimir
On Tue, Apr 19, 2016 at 1:29 PM, pari eghbali <[log in to unmask]> wrote:
> Dear Vladimir,
>
> Thanks a lot for your advise. I am going to give you a short summary of my
> analysis in order to have your comment about it. I inverted two DCMs
> corresponding to different hypothesis: 1- connectivity decrease across
> trials (M1) 2- connectivity increase across trials (M2) with same structure.
> Winner model is M1. According to your advise, I performed one sample ttest
> on just DCM.Ep.B (2,1) of two models across subjects in order to infer
> about parameter. Parameter of Model1 is significant p=0.002 but in order to
> report these results as % decrease in connectivity between source 1 to
> source 2 (100*exp(DCM.Ep.B(2,1)*decrease function)), I see all subjects'
> connectivity strength increases across trials. How Can I interpret my result
> since it is against the own model (connectivity decrease across trials).
> Another question is about one sample ttest, DCM.Pp.B(2,1) should not be
> considered in one sample ttest?
>
> Your advice will be really helpful again.
>
> Regards,
>
> Pari
>
>
> On Wed, Apr 13, 2016 at 4:42 PM, Vladimir Litvak <[log in to unmask]>
> wrote:
>>
>> Dear Pari,
>>
>> I don't think there is an easy way to force the direction of the effect to
>> be positive or negative. I would suggest that you test for direction by
>> doing a classical t-test on the B parameter across subjects.
>>
>> Best,
>>
>> Vladimir
>>
>> On Wed, Apr 13, 2016 at 3:39 PM, pari eghbali <[log in to unmask]>
>> wrote:
>>>
>>> Dear SPM experts,
>>> I defined two models: 1- Model defined based on between trial increase
>>> hypothesis and 2- Model defined based on between trial decrease hypothesis.
>>> Model1 is winner model (BMS) but its B matrix is negative and
>>> consequently connectivity decrease across trials!!! so by this way, winner
>>> Model hypothesis (between trial increase) which is in contrast of its own
>>> result is correct or no?
>>> I am totally confused about this phenomenon and I need help.
>>> Thank you very much for your response.
>>>
>>> Best,
>>> Pari
>>
>>
>
|