nimare.ml.describe_fields

describe_fields(studyset, source=None, min_coverage=0.0)[source]

Return the fields to_bunch() can read.

A release-scale Studyset offers hundreds of metadata columns and hundreds of annotation labels, most of which no analysis fills in. This reports what each one holds, using the same reader to_bunch() uses, so a field described as numeric here is numeric there.

Parameters:

  • studyset (Studyset) – The Studyset to describe.

  • source ({"metadata", "annotations", "texts"}, optional) – Restrict the report to one source, by default None, meaning all three.

  • min_coverage (float, optional) – Drop fields reported by a smaller fraction of analyses than this, by default 0.0, which keeps every field.

Returns:

One row per field, ordered by coverage, with columns source, field, kind, coverage, n_unique and example.

Return type:

pandas.DataFrame

Examples

>>> fields = describe_fields(studyset, min_coverage=0.5)
>>> fields[fields.kind == "numeric"].head()

Examples using nimare.ml.describe_fields

Machine learning in NiMARE

Machine learning in NiMARE