-
Notifications
You must be signed in to change notification settings - Fork 12
[WIP] GLM
Thomas Vincent edited this page Nov 13, 2018
·
2 revisions
The process NIRS - wip -> GLM - design and fit
takes care of building the design matrix from the stimulation events and fits the GLM on the input data.
It can handle data in the channel-space or projected on the cortical surface (TODO: see).
- Stimulation events: list of event group names for which to build stimulus-induced regressors.
-
HRF model: model to build the hemodynamic time course. Note that the HRF duration is fixed to 25 sec.
-
CANONICAL
: canonical HRF as in Glover 1995 -
GAMMA
: sum of two gamma functions -
BOXCAR
: binary boxcar / step function
-
- Add constant regressor: append a column of ones to the design matrix.
-
Fitting method:
-
OLS
: ordinary least square (white noise). -
AR-ILS
: autoregressive noise model. This method uses the Analyzir toolbox which must be installed prior to running this process.
-
- Extra outputs: export intermediary results.
A custom brainstorm object containing all GLM results is produced, named as <input data> | GLM <fit method> - results
. It exposes the design matrix, which can be viewed by double clicking on it.
Extra outputs (if enabled in parameters):
- Residuals: residual time-series after model fitting ( Y - X * B_fit )
- Beta maps: maps of estimated beta.
See process NIRS - wip -> GLM - intra subject contrast
.
The contrast computation takes the results of the previous process as input (<input data> | GLM <fit method> - results
)
-
Tutorials:
-
Workshop and courses:
-
Manual:
-
Contribute: