modalities.fmri.fmristat.model

Module: modalities.fmri.fmristat.model

Inheritance diagram for nipy.modalities.fmri.fmristat.model:

digraph inheritance6888041a7d { rankdir=LR; size="8.0, 12.0"; "fmristat.model.AR1" [URL="#nipy.modalities.fmri.fmristat.model.AR1",fontname="Vera Sans, DejaVu Sans, Liberation Sans, Arial, Helvetica, sans",fontsize=10,height=0.25,shape=box,style="setlinewidth(0.5)",target="_top",tooltip="Second pass through fmri_image."]; "fmristat.model.ModelOutputImage" [URL="#nipy.modalities.fmri.fmristat.model.ModelOutputImage",fontname="Vera Sans, DejaVu Sans, Liberation Sans, Arial, Helvetica, sans",fontsize=10,height=0.25,shape=box,style="setlinewidth(0.5)",target="_top",tooltip="These images have their values filled in as the model is fit, and"]; "fmristat.model.OLS" [URL="#nipy.modalities.fmri.fmristat.model.OLS",fontname="Vera Sans, DejaVu Sans, Liberation Sans, Arial, Helvetica, sans",fontsize=10,height=0.25,shape=box,style="setlinewidth(0.5)",target="_top",tooltip="First pass through fmri_image."]; }

This module defines the two default GLM passes of fmristat

The results of both passes of the GLM get pushed around by generators, which know how to get out the (probably 3D) data for each slice, or parcel (for the AR) case, estimate in 2D, then store the data back again in its original shape.

The containers here, in the execute methods, know how to reshape the data on the way into the estimation (to 2D), then back again, to 3D, or 4D.

It’s relatively easy to do this when just iterating over simple slices, but it gets a bit more complicated when taking arbitrary shaped samples from the image, as we do for estimating the AR coefficients, where we take all the voxels with similar AR coefficients at once.

Classes

AR1

class nipy.modalities.fmri.fmristat.model.AR1(fmri_image, formula, rho, outputs=[], volume_start_times=None)

Bases: object

Second pass through fmri_image.

Parameters

fmri_image : FmriImageList

object returning 4D array from np.asarray, having attribute volume_start_times (if volume_start_times is None), and such that object[0] returns something with attributes shape

formula : nipy.algorithms.statistics.formula.Formula

rho : Image

image of AR(1) coefficients. Returning data from rho.get_data(), and having attribute coordmap

outputs :

volume_start_times :

__init__(fmri_image, formula, rho, outputs=[], volume_start_times=None)

Initialize self. See help(type(self)) for accurate signature.

execute()

ModelOutputImage

class nipy.modalities.fmri.fmristat.model.ModelOutputImage(filename, coordmap, shape, clobber=False)

Bases: object

These images have their values filled in as the model is fit, and are saved to disk after being completely filled in.

They are saved to disk by calling the ‘save’ method.

The __getitem__ and __setitem__ calls are delegated to a private Image. An exception is raised if trying to get/set data after the data has been saved to disk.

__init__(filename, coordmap, shape, clobber=False)

Initialize self. See help(type(self)) for accurate signature.

save()

Save current Image data to disk

OLS

class nipy.modalities.fmri.fmristat.model.OLS(fmri_image, formula, outputs=[], volume_start_times=None)

Bases: object

First pass through fmri_image.

Parameters

fmri_image : FmriImageList or 4D image

object returning 4D data from np.asarray, with first (object[0]) axis being the independent variable of the model; object[0] returns an object with attribute shape.

formula : nipy.algorithms.statistics.formula.Formula

outputs :

volume_start_times :

__init__(fmri_image, formula, outputs=[], volume_start_times=None)

Initialize self. See help(type(self)) for accurate signature.

execute()

Functions

nipy.modalities.fmri.fmristat.model.estimateAR(resid, design, order=1)

Estimate AR parameters using bias correction from fMRIstat.

Parameters

resid: array-like

residuals from model

model: an OLS model used to estimate residuals

Returns

output : array

shape (order, resid

nipy.modalities.fmri.fmristat.model.generate_output(outputs, iterable, reshape=<function <lambda>>)

Write out results of a given output.

In the regression setting, results is generally going to be a scipy.stats.models.model.LikelihoodModelResults instance.

Parameters

outputs : sequence

sequence of output objects

iterable : object

Object which iterates, returning tuples of (indexer, results), where indexer can be used to index into the outputs

reshape : callable

accepts two arguments, first is the indexer, and the second is the array which will be indexed; returns modified indexer and array ready for slicing with modified indexer.

nipy.modalities.fmri.fmristat.model.model_generator(formula, data, volume_start_times, iterable=None, slicetimes=None, model_type=<class 'nipy.algorithms.statistics.models.regression.OLSModel'>, model_params=<function <lambda>>)

Generator for the models for a pass of fmristat analysis.

nipy.modalities.fmri.fmristat.model.output_AR1(outfile, fmri_image, clobber=False)

Create an output file of the AR1 parameter from the OLS pass of fmristat.

Parameters

outfile :

fmri_image : FmriImageList or 4D image

object such that object[0] has attributes coordmap and shape

clobber : bool

if True, overwrite previous output

Returns

regression_output : RegressionOutput instance

nipy.modalities.fmri.fmristat.model.output_F(outfile, contrast, fmri_image, clobber=False)

output F statistic images

Parameters

outfile : str

filename for F contrast image

contrast : array

F contrast matrix

fmri_image : FmriImageList or Image

object such that object[0] has attributes shape and coordmap

clobber : bool

if True, overwrites previous output; if False, raises error

Returns

f_reg_out : RegressionOutput instance

Object that can a) be called with a results instance as argument, returning an array, and b) accept the output array for storing, via obj[slice_spec] = arr type slicing.

nipy.modalities.fmri.fmristat.model.output_T(outbase, contrast, fmri_image, effect=True, sd=True, t=True, clobber=False)

Return t contrast regression outputs list for contrast

Parameters

outbase : string

Base filename that will be used to construct a set of files for the TContrast. For example, outbase=’output.nii’ will result in the following files (assuming defaults for all other params): output_effect.nii, output_sd.nii, output_t.nii

contrast : array

F contrast matrix

fmri_image : FmriImageList or Image

object such that object[0] has attributes shape and coordmap

effect : {True, False}, optional

whether to write an effect image

sd : {True, False}, optional

whether to write a standard deviation image

t : {True, False}, optional

whether to write a t image

clobber : {False, True}, optional

whether to overwrite images that exist.

Returns

reglist : RegressionOutputList instance

Regression output list with selected outputs, where selection is by inputs effect, sd and t

Notes

Note that this routine uses the corresponding output_T routine in outputters, but indirectly via the TOutput object.

nipy.modalities.fmri.fmristat.model.output_resid(outfile, fmri_image, clobber=False)

Create an output file of the residuals parameter from the OLS pass of fmristat.

Uses affine part of the first image to output resids unless fmri_image is an Image.

Parameters

outfile :

fmri_image : FmriImageList or 4D image

If FmriImageList, needs attributes volume_start_times, supports len(), and object[0] has attributes affine, coordmap and shape, from which we create a new 4D coordmap and shape If 4D image, use the images coordmap and shape

clobber : bool

if True, overwrite previous output

Returns

regression_output :

nipy.modalities.fmri.fmristat.model.results_generator(model_iterable)

Generator for results from an iterator that returns (index, data, model) tuples.

See model_generator.