Class: DecodingDetails¶

Payload for a classification or pattern-discriminability analysis. Above-chance decoding is one signed cell with nothing crossed, and the reference it beat is PerformanceMetric.reference_value, which is per metric because accuracy against chance and AUC against 0.5 are two references. A between-cohort accuracy comparison is a crossed cohort term; accuracy regressed on a behavioural score is a cell on a continuous term. Spatial scope is recorded on the Analysis, including for searchlight analyses.

__

"One signed cell" means exactly one. The class the reported accuracy is for takes the positive cell and the other class takes no cell at all, because absence is the zero weight. Signing both would say the paper reported which way the classifier erred, which a single accuracy does not. Where the test is undirected -- accuracy against chance, with no class named as the target -- every cell is undirected instead.

classes is then the only field saying what was told apart, so leaving it empty asserts that the paper never named the classes it decoded, which is rare.

URI: neurostudy_ex:DecodingDetails

 classDiagram
    class DecodingDetails
    click DecodingDetails href "../DecodingDetails/"
      AnalysisDetails <|-- DecodingDetails
        click AnalysisDetails href "../AnalysisDetails/"

      DecodingDetails : classes





        DecodingDetails --> "*" DecodingClass : classes
        click DecodingClass href "../DecodingClass/"



      DecodingDetails : decoded_variable





        DecodingDetails --> "1" ExtractedString : decoded_variable
        click ExtractedString href "../ExtractedString/"



      DecodingDetails : details_type

      DecodingDetails : generalization





        DecodingDetails --> "0..1" ExtractedString : generalization
        click ExtractedString href "../ExtractedString/"



      DecodingDetails : performance_metrics





        DecodingDetails --> "1..*" PerformanceMetric : performance_metrics
        click PerformanceMetric href "../PerformanceMetric/"



      DecodingDetails : validation_scheme





        DecodingDetails --> "0..1" ExtractedString : validation_scheme
        click ExtractedString href "../ExtractedString/"

Inheritance¶

AnalysisDetails
- DecodingDetails

Slots¶

Name	Cardinality and Range	Description	Inheritance
decoded_variable	1 ExtractedString	The variable the classifier or discriminability test predicted	direct
classes	* DecodingClass	The classes discriminated	direct
performance_metrics	1..* PerformanceMetric	The performance measures reported for this map, one entry each	direct
validation_scheme	0..1 ExtractedString	Cross-validation or resampling scheme, such as leave-one-run-out	direct
generalization	0..1 ExtractedString	Train/test split for cross-decoding, when the analysis generalized across set...	direct
details_type	1 String	The concrete details subclass carried by this payload	AnalysisDetails

Identifier and Mapping Information¶

Schema Source¶

from schema: https://neurostuff.org/schemas/neuroimaging-study-extraction

Mappings¶

Mapping Type	Mapped Value
self	neurostudy_ex:DecodingDetails
native	neurostudy_ex:DecodingDetails

LinkML Source¶

Direct¶

name: DecodingDetails
description: 'Payload for a classification or pattern-discriminability analysis. Above-chance
  decoding is one signed cell with nothing crossed, and the reference it beat is `PerformanceMetric.reference_value`,
  which is per metric because accuracy against chance and AUC against 0.5 are two
  references. A between-cohort accuracy comparison is a crossed cohort term; accuracy
  regressed on a behavioural score is a cell on a continuous term. Spatial scope is
  recorded on the Analysis, including for searchlight analyses.


  "One signed cell" means exactly one. The class the reported accuracy is *for* takes
  the positive cell and the other class takes no cell at all, because absence is the
  zero weight. Signing both would say the paper reported which way the classifier
  erred, which a single accuracy does not. Where the test is undirected -- accuracy
  against chance, with no class named as the target -- every cell is `undirected`
  instead.

  `classes` is then the only field saying what was told apart, so leaving it empty
  asserts that the paper never named the classes it decoded, which is rare.'
from_schema: https://neurostuff.org/schemas/neuroimaging-study-extraction
rank: 1000
is_a: AnalysisDetails
attributes:
  decoded_variable:
    name: decoded_variable
    description: The variable the classifier or discriminability test predicted.
    from_schema: https://neurostuff.org/schemas/neuroimaging-study-extraction/analysis_details
    rank: 1000
    domain_of:
    - DecodingDetails
    range: ExtractedString
    required: true
  classes:
    name: classes
    description: 'The classes discriminated. These are non-directional: direction,
      where it exists, is carried by Effect.'
    from_schema: https://neurostuff.org/schemas/neuroimaging-study-extraction/analysis_details
    rank: 1000
    domain_of:
    - DecodingDetails
    range: DecodingClass
    multivalued: true
    inlined_as_list: true
  performance_metrics:
    name: performance_metrics
    description: The performance measures reported for this map, one entry each. Papers
      routinely give several of the same analysis -- accuracy with sensitivity and
      specificity, or accuracy with AUC -- and each carries its own value and its
      own reference. Record every one the source reports.
    from_schema: https://neurostuff.org/schemas/neuroimaging-study-extraction/analysis_details
    rank: 1000
    domain_of:
    - DecodingDetails
    range: PerformanceMetric
    required: true
    multivalued: true
    inlined_as_list: true
  validation_scheme:
    name: validation_scheme
    description: Cross-validation or resampling scheme, such as leave-one-run-out.
    from_schema: https://neurostuff.org/schemas/neuroimaging-study-extraction/analysis_details
    rank: 1000
    domain_of:
    - DecodingDetails
    range: ExtractedString
  generalization:
    name: generalization
    description: Train/test split for cross-decoding, when the analysis generalized
      across sets.
    from_schema: https://neurostuff.org/schemas/neuroimaging-study-extraction/analysis_details
    rank: 1000
    domain_of:
    - DecodingDetails
    range: ExtractedString

Induced¶

name: DecodingDetails
description: 'Payload for a classification or pattern-discriminability analysis. Above-chance
  decoding is one signed cell with nothing crossed, and the reference it beat is `PerformanceMetric.reference_value`,
  which is per metric because accuracy against chance and AUC against 0.5 are two
  references. A between-cohort accuracy comparison is a crossed cohort term; accuracy
  regressed on a behavioural score is a cell on a continuous term. Spatial scope is
  recorded on the Analysis, including for searchlight analyses.


  "One signed cell" means exactly one. The class the reported accuracy is *for* takes
  the positive cell and the other class takes no cell at all, because absence is the
  zero weight. Signing both would say the paper reported which way the classifier
  erred, which a single accuracy does not. Where the test is undirected -- accuracy
  against chance, with no class named as the target -- every cell is `undirected`
  instead.

  `classes` is then the only field saying what was told apart, so leaving it empty
  asserts that the paper never named the classes it decoded, which is rare.'
from_schema: https://neurostuff.org/schemas/neuroimaging-study-extraction
rank: 1000
is_a: AnalysisDetails
attributes:
  decoded_variable:
    name: decoded_variable
    description: The variable the classifier or discriminability test predicted.
    from_schema: https://neurostuff.org/schemas/neuroimaging-study-extraction/analysis_details
    rank: 1000
    owner: DecodingDetails
    domain_of:
    - DecodingDetails
    range: ExtractedString
    required: true
  classes:
    name: classes
    description: 'The classes discriminated. These are non-directional: direction,
      where it exists, is carried by Effect.'
    from_schema: https://neurostuff.org/schemas/neuroimaging-study-extraction/analysis_details
    rank: 1000
    owner: DecodingDetails
    domain_of:
    - DecodingDetails
    range: DecodingClass
    multivalued: true
    inlined: true
    inlined_as_list: true
  performance_metrics:
    name: performance_metrics
    description: The performance measures reported for this map, one entry each. Papers
      routinely give several of the same analysis -- accuracy with sensitivity and
      specificity, or accuracy with AUC -- and each carries its own value and its
      own reference. Record every one the source reports.
    from_schema: https://neurostuff.org/schemas/neuroimaging-study-extraction/analysis_details
    rank: 1000
    owner: DecodingDetails
    domain_of:
    - DecodingDetails
    range: PerformanceMetric
    required: true
    multivalued: true
    inlined: true
    inlined_as_list: true
  validation_scheme:
    name: validation_scheme
    description: Cross-validation or resampling scheme, such as leave-one-run-out.
    from_schema: https://neurostuff.org/schemas/neuroimaging-study-extraction/analysis_details
    rank: 1000
    owner: DecodingDetails
    domain_of:
    - DecodingDetails
    range: ExtractedString
  generalization:
    name: generalization
    description: Train/test split for cross-decoding, when the analysis generalized
      across sets.
    from_schema: https://neurostuff.org/schemas/neuroimaging-study-extraction/analysis_details
    rank: 1000
    owner: DecodingDetails
    domain_of:
    - DecodingDetails
    range: ExtractedString
  details_type:
    name: details_type
    description: The concrete details subclass carried by this payload. Populated
      from the class itself, not extracted.
    from_schema: https://neurostuff.org/schemas/neuroimaging-study-extraction/analysis_details
    rank: 1000
    designates_type: true
    owner: DecodingDetails
    domain_of:
    - AnalysisDetails
    range: string
    required: true