Further STATO reuse: Design Matrix and Z statistic

[cmaumet]

I propose the following actions

• nidm:DesignMatrix : Replace with external term "design matrix" http://purl.obolibrary.org/obo/STATO_0000289 (in fact proposed by Nolan!)
• nidm:ZStatistic : This term can be deleted. (SPM & FSL have nidm:equivalentZStatistic on the results page). The term "z-score" can also be directly references: http://purl.obolibrary.org/obo/STATO_0000104

Current definitions:

• nidm:DesignMatrix : A matrix of values defining the explanatory variables used in a regression model. Each column corresponds to one explanatory variable, each row corresponds to one observation
• nidm:ZStatistic : undefined

@jbpoline , @cmaumet: Comments?

(See PURL links for STATO definitions)

[cmaumet]

(Following the merge of PR #262, re-use of stato:Z-statistic is now implemented.)

Regarding nidm:DesignMatrix, one thing we need to think about is whether we want to deal with spatially varying design matrices (under discussion at #207). If so, we might want to have a definition that refer to the stato:DesignMatrix (similarly to what was done for nidm:ModelParametersEstimation in #141, cf. this comment) but is not strictly equivalent to it. Do you see what I mean?

[nicholst]

OK, I'm fine with keeping nidm:DesignMatrix if it gives us greater flexibility later to deal with spatially designs. However, that's a rare enough use case I won't be upset if we don't explicitly represent it now.

[cmaumet]

+1 to re-use directly stato:'design matrix' (and keep the spatially varying design matrix for a later extension).

[nicholst]

Upon discussion with @cmaumet, we think we should not reuse the stato:'design matrix' term. The reasons is that our nidm:DesignMatrix entity can take some very fMRI-specific attributes, like HRF and drift descriptions.

We propose a definition that closely follows and references the STATO definition, however:

• nidm:DesignMatrix: A "design matrix" as defined by http://purl.obolibrary.org/obo/STATO_0000289 , with additional neuroimaging attributes, including HRF and drift for first level fMRI models.

[khelm]

+1 , sounds good to me.

[cmaumet]

Thank you! Now implemented and ready for release.