This is the LAST release of Nistats.

All Nistats functionality has been incorporated into Nilearn.
It will be available as part of Nilearn 0.7.0.
All future development will be done in Nilearn.



The phrases resid and w in method and attribute names
have been changed to residuals and whitened_ respectively.
This will be a breaking change in the stable release of Nistats,
due to be shipped with Nilearn 0.7.0 as nilearn.stats.
To minimize compatibility problems, please change the names
according to the info in the “Changes” section.



  • The following parameters, methods, and attributes have been renamed. Classes affected: OLSModel, RegressionResults, LikelihoodModelResults.

    • resid -> residuals

    • df_resid -> df_residuals

    • norm_resid -> normalized_residuals

    • wresid -> whitened_residuals

    • wdesign -> whitened_design

    • wX -> whitened_X

    • wY -> whitened_y

  • run_img variable deleted after masking in FirstLevelModel to reduce memory use.

  • nistats.reporting.make_glm_report now uses the computed mask if there is no user-specified one.

  • Default value of alpha in nistats.reporting.make_glm_report is 0.001.


  • Explicit method for fixed effects to support image-based summary statistics approach.

  • FIR delays are now integers.



period_cut (in seconds) has been replaced by high_pass (in Hz) in FirstLevelModel & design matrices.
This is a code-breaking change. Please update your code accordingly.



  • Nistats now uses BIDS v1.2 & BIDS Derivatives terminology.



The following people contributed to this release (in alphabetical order):

Ana Luisa Pinho
Anthony Gifuni
Bertrand Thirion
Christopher J. Markiewicz
Christophe Pallier
Gael Varoquaux
Jerome Dockes
Jerome-Alexis Chevalier
Jessica Thompson
Kshitij Chawla (kchawla-pi)
Takis Panagopoulos
Tuan Binh Nguyen



  • Updated the minimum versions of the dependencies
    • Numpy >= 1.11

    • SciPy >= 0.17

    • Nibabel >= 2.0.2

    • Nilearn >= 0.4.0

    • Pandas >= 0.18.0

    • Sklearn >= 0.18.0

  • Added comprehensive tutorial

  • Second-level model accepts 4D images as input.

  • Changes in function parameters
    • third argument of map_threshold is now called level.

    • Changed the defaut oversampling value for the hemodynamic response to 50 and exposed this parameter.

    • changed the term paradigm to events and made it BIDS-compliant. Set the event file to be tab-separated

    • FirstLevelModel.compute_contrasts parameter output_type can take the value 'all', returning a dictionary of images for each output type

  • Certain functions and methods have been renamed for clarity
    • nistats.design_matrix
      • make_design_matrix() -> make_first_level_design_matrix()

      • create_second_level_design() -> make_second_level_design_matrix()

    • nistats.utils
      • pos_recipr() -> positive_reciprocal()

      • multiple_fast_inv() -> multiple_fast_inverse()

  • Python2 Deprecation:

    Python 2 is now deprecated and will not be supported in a future version. A DeprecationWarning is displayed in Python 2 environments with a suggestion to move to Python 3.


The following people contributed to this release:

45  Bertrand Thirion
70  Kshitij Chawla
16  Taylor Salo
 6  KamalakerDadi
 5  chrplr
 5  hcherkaoui
 5  rschmaelzle
 4  mannalytics
 3  Martin Perez-Guevara
 2  Christopher J. Markiewicz
 1  Loïc Estève



First alpha release of nistats.

Contributors (from git shortlog -ns):

223  Martin Perez-Guevara
195  bthirion
 24  Gael Varoquaux
  9  Loïc Estève
  3  AnaLu
  2  Alexandre Gramfort
  1  DOHMATOB Elvis
  1  Horea Christian
  1  Michael Hanke
  1  Salma
  1  chrplr