Hi,
> Could you please let me know the answers? Your help is greatly appreciated. Thank you.
>
> 1. p.20 "An example experiment An FMRI adaptation of a classical PET experiment": What is 6 sec ISA?
It is a 6 second spacing between events.
> 2. p.30 "Building a model": “This time we used the onset times for the shadowing events to get the predicted brain response for those.” Could you please let us know the missing the ending part?
It is complete - the "those" refers to the shadowing events.
> 3. p.31 “Building a model”. “And we can look for voxels that match that”. What is the missing the ending part?
This is also complete - the "that" refers to the predicted response for the shadowing events (the last thing to be described on the previous slide).
> 4. p.33 “Slight detour: Making regressors”. What is mean by “Sub-sample at Tr of experiment”?
Sub-sample = down-sample. Tr = TR = new sampling rate.
> 5. p.49 "t-contrasts" There are three red arrows from "The Model", "The Contrast" and "and the Residual Error". By "Depends on", do they mean std(COPE) depends on these three factors?
Yes.
> 6. p.49-50 How to go from t, COPE and std(COPE) to the brain images in the next slide?
Each image represents the corresponding quantity (each voxel contains that quantity).
> 7. p.49 Residual error = standard error = standard derivation?
No. Standard error and standard deviation are different. The residual error is the same as the standard deviation in this case.
> 8. p.61 "t-contrasts": the author asks "Why [1 –1] instead of [1 0]?" So why?
This makes sense in the context of the talk, but I wouldn't worry about it here.
> 10. p.69 "F-contrasts": it states that F = (Ssr-Sse)/Sse = (large arrow – small arrow / small arrow). What do those double arrows of different lengths mean?
They represent amplitudes.
> 11. p.69 "F-contrast": What is Reduced Model? Why the blue line next to the reduced model's matrix is flat?
There is nothing left in the reduced model, so the blue line is just zero.
> 12. p.71 "F-contrast vs Mean activation": Why "Most probably not significant"?
Because the mean activation value (-0.1) is small.
> 13. p.81 In "Choosing High-Pass Filter Cut-off", under "Example: Boxcar EV with period 100s" "Negligible effect on EV, so use cut-off of 100s" Could you please let me know why negligible and why use cut-off of 100s?
>
> 14. Same page under "Example: Boxcar with period 250s" "Substantial effect on EV" Could you please explain why substantial effect and why a cut-off of 100s is chosen?
>
> 15. Similarly, under the rightmost design matrix, it states "Negligible effect on EV, so use cut-off of 250s". Please let us know why negligible and why use a cut-off of 250s.
These are all to do with whether the shape of the waveform (the boxcar) is distorted by the filter or not, as distortion means that signal is being removed which is bad.
> 16. p.82 and p.84, 85: What is autocorrelation? Can't find it in fMRI books.
I'm surprised by this, as it is not just FSL that deals with autocorrelation. It is a general mathematical concept though - so look it up elsewhere. The internet has a lot of information...
> 17. p.84 FMRIB's Improved Linear Modelling (FILM): Could you please explain what "Apply filter to data and design matrix and refit" is?
The pre-whitening filter is applied to both sides of the GLM equation (data and model).
All the best,
Mark
|