bthirion
Currently, a string can only a design matrix column if it fits the `isidentifier()` condition, which looks excessive. For instance, strings started with a number are not accepted, while there...
`permuted_ols` actually works for one-sample tests, which is not documented. Also, the outputs are only -log_10( corrected_p_values), which is a bit limited too. It is also ill-documented. ### Affected pages
The example demoes the advantage of non-parametric vs parametric inference, but it conflates it with a difference in control type: fwer vs fdr. The parametrix and non-parametric methods should be...
The funding support information displayed in https://nilearn.github.io/authors.html#funding is outdated and should be updated. ### Affected pages https://nilearn.github.io/authors.html#funding
Stores un-necessary stuff (input data, residuals), hence memory-consuming. Partially handled in nilearn/nistats#72, but full removal should be considered.
Naturally comes after nilearn/nistats#26 can be mostly copy-pasted from nipy.
I encounter some issues using fetch_neurovault: * the filtering does not seem to work: the images returned do not have the expected 'cognitive_paradigm_cogatlas_id' * it downloads many images -- more...
Should introduce optional possibility to orthogonalize regressors (parametric modulation).
Addresses #3194. Still need more tests (for F tests). Detecting some inconsistencies in dimension: se sometimes require effect sizes for t contrasts to have shape (1, n_voxels) while it would...
Curenntly `computes_fixed _effects` performs 'manual' computation to go from effects and variance to statsitics, implictly assuming a t stat. This is bad. What we should do would be to reinstantiate...