Re: RSA Analysis Inquiry

[Nick Oosterhof]



Greetings,

On 28 Feb 2024, at 17:47, 'Olivia Yang' via CoSMoMVPA <cosm...@googlegroups.com> wrote:

In our experiment, we presented six face stimuli to subjects using a block design, with two emotions (neutral and angry) across three identities. After scanning, subjects were asked to rate the trustworthiness of each face stimulus on a scale from 0 to 10. Our goal is to utilize RSA to investigate two main aspects: firstly, to uncover any shared mechanisms in the representation of face emotion and trustworthiness, and secondly, to explore whether the representation of face emotion influences the representation of trustworthiness.

Can CosMoMVPA effectively accomplish both of the goals? I think we can calculate the neural RDM using the emotion category RDM first, followed by calculating the neural RDM using the rating RDM.

I’m a bit confused by this description. I could think of these possible RDMs:

1. A neural RDM based on fMRI data from a region of interest.

2. An emotion behavioural RDM based on similarity in emotion.

3. A trustworthiness behavioural RDM based on trustworthiness judgments.

I am not sure what you mean by “ calculating the neural RDM using the rating RDM”

RSA could (at least in theory) be done by correlating the neural RDM with either of the behavioural RDMs. Or through regression analysis using both behavioural RDMs.

Then I plan to conduct a regression analysis between these two neural RDMs. Does this way seem appropriate?

In theory it may be possible, but I see some challenges: 

- there are few conditions.

- the emotion and trustworthiness aspect are most likely not independent. (This may seem like shameless self promotion, but quite a few papers of which I’m a co-author support this notion). 

Together this could make it hard to find significant effects for the analyses you seem to be interested in.

Best,

Nick