Forum

Notifications
Clear all

Help with beta extraction

7 Posts
5 Users
0 Likes
583 Views
Bruna Baldasso
Posts: 11
Topic starter
(@bdalcinbalda)
Active Member
Joined: 1 year ago

Hello,

I have a question about the beta coefficients shown in the output. 

I used a GLM with the Gaussian basis function, and I now have 6 different beta coefficients, which I read that are one for each basis function. How do I choose between those beta? (see picture) They seem quite different from one another. How do I get to the final beta?

Thank you very much,

Bruna

 

 

6 Replies
David Boas
Posts: 238
(@dboas)
Reputable Member
Joined: 2 years ago

We use consecutive gaussian basis functions to model the HRF to allow for more flexiblity in the estimated shape of the HRF. This is quite different than using, for instance, a gamma function with specified width and delay parameters and then you estimate the beta which is just the amplitude. In the gamma function case, you get a single beta coefficient for the amplitude that you can do your statistics on. In the gaussian basis functions case, you need to multiple the series of beta coefficients by the basis functions to restore the shape of the HRF. Then you can measure the mean across a chosen time range of that HRF. Then you can look at the statistics of that mean across subjects.

Reply
1 Reply
Janine Camatti
(@janine-cmatti)
Joined: 4 months ago

New Member
Posts: 1

@dboas Hi David,

 
I came across the same problem. It seems a little bit more clear after your explanation, but I still have a few doubts. For instance, after run the GLM using the consecutive gaussian basis function I ended up with 18 beta values for each condition, from this, how do I use this values to estimate the HRF?  The dcAvg output gives me the reconstructed HRF, is that correct? If so, I can perform the statistics comparing the dcAvg across subjects?
And for last, just a technical issue: In the GLM function I set the trange at -2 and 35, but for the group display it only show a range of -2 and 20. I could not find where to set the group HFR length right.
 
Thanks!
 
Reply
Bruna Baldasso
Posts: 11
Topic starter
(@bdalcinbalda)
Active Member
Joined: 1 year ago

Hello David,

Thank you very much, that makes sense. Can you just explain how to do this on Matlab for someone with very basic calculus background and little matlab skills? 

"you need to multiple the series of beta coefficients by the basis functions to restore the shape of the HRF. Then you can measure the mean across a chosen time range of that HRF. Then you can look at the statistics of that mean across subjects."

Also, if I choose to use the gamma function, do you recommend using the default parameters of [0.1 3.0 10.0 1.8 3.0 10.0]? Sorry I am not very familiar with this function.

I really appreciate your help!!

 

 

Reply
Meryem Yücel
Posts: 133
Admin
(@mayucel)
Estimable Member
Joined: 2 years ago

Hi Bruna,

 

That multiplication results in HRF, and Homer already outputs HRF with the output name "dcAvg". Please check out the link below. You can also export HRF.

https://github.com/BUNPC/Homer3/wiki/Output-Files

 

Meryem

Reply
Bruna Baldasso
Posts: 11
Topic starter
(@bdalcinbalda)
Active Member
Joined: 1 year ago

Hi @dboas,

I ended up using the gamma function like you had mentioned. 

I was just wondering if there is a better way of exporting the betas in Matlab? I am only able to see them when I go in each subject's .mat file. Is there a way to get all betas for all subjects and conditions?

Thank you!

Reply
1 Reply
Neta Weitzman
(@neta-weitzman)
Joined: 1 year ago

Eminent Member
Posts: 8

Hey @bdalcinbalda,

Did you find an easier way to extract each subject betas?

Thanks!

Reply
Share:
en_USEnglish