OverlappedLinearModel¶
Estimate a linear model with temporally overlapped responses.
The model applies a GLM that explains the observed signal as a sum of stimulus-locked responses, (according one of the selectable response models) for different stimulus instance types, plus optional time-varying nuisance regressors from selectable signal chunks, plus noise. The model controls for the false discovery rate due to temporally correlated observations by spectrally whitening the data, and for artifacts by using a robust (iteratively reweighted least squares) estimator. When using the FIR model, the response is estimated in time-series form, and when using the Canonical HRF model, the magnitude of the modeled response is estimated (i.e., not temporally resolved). More Info... Version 0.9.0
Ports/Properties¶
data¶
Data to process.
response_model¶
Evoked response model. This is the model of any event-locked responses expected in the data. FIR will estimate the response as a time series. Canonical HRF will estimate the amplitude of a canonical hemodynamic response function. The response is time-locked to the onset and duration of instances in a marker stream.
response_parameters¶
Parameter overrides for the response model, given as a Python dictionary. The parameters are per the documentation of the CanonicalHRF and FIRHRF models, respectively. In brief. FIRHRF supports numBins, binWidth, binDelay, and timeUnit parameters, and CanonicalHRF supports
max_autocorr_delay¶
Estimate auto-correlations up to at most this delay (in seconds). This translates to the maximum model order of an autoregressive model that is being estimated alongside the linear response model. The actual model order may be lower when it is estimated from the data. If set to 0, no autocorrelations are accounted for.
robust¶
Use robust estimation for linear model. This uses an iteratively reweighted least-squares (IRLS) method using the Tukey Bisquare scale function.
regressor_streams¶
Streams with time-domain nuisance or auxiliary regressors. This can be additional streams that carry explanatory variables that shall be regressed out of the data as part of the model, for instance auxiliary measures of artifacts, or certain kinds of low-order trend terms etc (e.g., via the node AppendSyntheticChannels, which can add such a stream).
statistic_axis¶
Output descriptive statistics along an axis of this type.
dependent_streams¶
List of stream names to use as dependent variables (default: all remaining).
max_iter¶
Max number of outer-loop (AR refit) iterations in optimization.
max_iter_inner¶
Max number of inner-loop (IRLS) iterations in optimization.
reltol¶
Max relative error during estimation.
use_caching¶
Enable caching.
num_procs¶
Number of parallel processes to use. The maximum would be the number of channels in the data. If given as -1, this will use one process per core on the system, and if given as None, this will use as many processes as there are cores in the system, divided by OMP_NUM_THREADS (or 4 if not present).
set_breakpoint¶
Set a breakpoint on this node. If this is enabled, your debugger (if one is attached) will trigger a breakpoint.
metadata¶
User-definable meta-data associated with the node. Usually reserved for technical purposes.