Email discussion lists for the UK Education and Research communities

## SPM@JISCMAIL.AC.UK

#### View:

 Message: [ First | Previous | Next | Last ] By Topic: [ First | Previous | Next | Last ] By Author: [ First | Previous | Next | Last ] Font: Proportional Font

#### Options

Subject:

Re: different DCM results between simple averaging and BPA averaging

From:

Zeidman, Peter

Date:

Wed, 15 Feb 2017 14:08:50 +0000

Content-Type:

text/plain

Parts/Attachments:

 text/plain (1 lines)
 ```Dear Toshi The simplest way to test this hypothesis, if you were starting at the beginning of the DCM analysis, would be to have a single parameter in your model representing the difference between your two experimental conditions. E.g. you would form a contrast [1 -1] in your GLM, and then use that as a modulatory input in the DCM. You could then do BPA and look in BPA.Pp for the posterior probability that the posterior estimate is non-zero. However, I infer that you have your the conditions encoded as two separate modulatory inputs entering your DCM, which you'd like to compare. Here are two ways of addressing your question: Option 1: Model comparison ======================== 1. I'll assume you are starting with just one DCM per subject. Create separate DCMs for each subject with both conditions as inputs, each individual condition as an input, and neither condition as input (null model) 2. Perform Bayesian Model Selection to select the best of these DCMs Option 2: Parameter comparison ========================== 1. Compute the BPA over subjects and save as DCM_XX.mat 2. In the GUI, click Dynamic Causal Modelling, then click Review, and select the DCM you created above. 3. Click Contrast of connections. Enter a contrast to compare condition A parameters against condition B parameters. E.g. 1s for all condition A parameters and -1s for all condition B parameters. (Note that the contrast must sum to zero.) This will give you the probability of a difference between conditions. Note that Option 1 is better in principle, because the model comparison takes into account the full covariance between parameters. Best Peter -----Original Message----- From: Toshi Kawagoe [mailto:[log in to unmask]] Sent: 15 February 2017 13:52 To: Zeidman, Peter <[log in to unmask]> Subject: Re: different DCM results between simple averaging and BPA averaging Dear Peter Yes, what I would like to do is to know the parameters and compare two condition within-subject design. There are two condition for each participants and I am trying to demonstrate that specific connectivity is stronger in one condition than in the other condition. Is the t-test enough to indicate those things? Best regards, Toshi ```

#### JiscMail Tools

JiscMail is a Jisc service.

View our service policies at https://www.jiscmail.ac.uk/policyandsecurity/ and Jisc's privacy policy at https://www.jisc.ac.uk/website/privacy-notice