Dear Darren,
At 18:54 15/11/2005, Darren Gitelman wrote:
>Dear Klaas, etc.
>
>I know Tali sent a follow-up question, but there were some specific
>confusing points in your answer (to me anyway) that I wanted to clarify.
>
>At 01:35 PM 11/14/2005, Klaas Enno Stephan wrote:
>>Dear Tali,
>>
>>it is usually not a good idea to use the same input as a driving
>>input to one region and as a modulator of an afferent connection
>>originating in that region.
>
>Do you mean afferent connection TO the driving region, or efferent
>connections FROM the driving region or both?
The latter.
>>The reason is that the conditional estimates of both parameters
>>will be highly correlated and, given that the prior variance for
>>driving inputs is higher than for modulatory ones, this will tend
>>to over-estimate the driving parameters and under-estimate the
>>modulatory parameters. In other words, your estimates are
>>based. (In complex networks this is not always true, but it is a
>>good rule-of-thumb. Individual cases are best checked using simulations.)
>
>OK. But it seems that the overestimates of driving effects would be
>largely confined to the connections emanating from the driving
>regions and not to modulatory effects elsewhere. Nevertheless, your
>point is well taken.
Some time ago I have looked at this question (i.e. the effects of
identical inputs) in simulations and found that one can still find
effects elsewhere in the network. They tend to be weaker the farther
they are away from the region where the input and the modulation of
its outgoing connection are identical, but the effects can vary
depending on what kind of network you examine exactly. There are
also types of networks where using identical inputs at multiple sites
in the network do not necessarily have problematic effects or only
for some regions. I have a paper in the pipeline addressing such
issues using simulations.
>>In your case, it seems best to define an input that includes words
>>and symbols (assuming that they equally activate the input region,
>>let's call it A1) but not fixation (assuming that this does not
>>activate A1) and define a second, modulatory input that only includes words.
>
>I think in Tali's design fixation was a distinct period in which
>there was a fixation point, and it was separate from Null events, so
>she may want to model it as well.
>
>> In this way, you are testing whether a word-related activation
>> in the target region (A2) can be explained by a selective increase
>> of the connection from A1 to A2 during the presentation of words.
>
>In addition I suppose one could also include the symbols effect as
>modulatory so that one had the following model
If the target region shows an effect of symbols as well, then
yes. Note that all that DCM does is to "re-explain" local
activations (main effects, interactions in your SPMs) in terms of
perturbations, connections and their modulation by experimental
factors. If there are no local activation elicited by an
experimental manipulation, DCM will not reveal anything magic about
connectivity.
>Let's say we have 3 regions. A1 gets driving inputs, and the network
>is fully interconnected, A1 <-> A2, A1 <-> A3, A2 <-> A3.
>Driving inputs are the combined vector of onsets for both words and symbols .
>
>Modulatory vectors are words and symbols separately, each of which
>effects connections A2 -> A3, and A3 -> A2, but are not modeled as
>affecting connections to or from A1.
>
>I think this would test selective increases of connection A2 -> A3
>or vice versa, during presentation of words, during presentation of
>symbols or one could calculate their difference. What do you think?
This looks fine to me. However, by defining the inputs in this way
(they will be correlated to some degree but far from showing perfect
correlation) you should also be able to modulate the A1->A2
connection. See the example model of the visual system in Karl's DCM
paper (2003 Neuroimage) where the situation is comparable (driving
input = static + moving + attended stimuli, modulatory input = moving stim)
Best wishes
Klaas
|