NAMIC Wiki:ShapeAnalysis:OutputFeature format

Shape Analysis Output of Statististical Tests

This page serves as a main starting point for discussion regarding describing the result from statistical tests on shape data.

General

• 2 main statistical tests: Global tests and local tests
• Results of these tests can be cumulative for the whole study (such as p-values) or population (such as an average object) or per object (such as a classification)
  • There is a great deal of flexibility for each filter, but the filter needs to store its results in the studyObject datastructure in a consistent way

Global Tests/Data

Global tests return single scalar values, such as p-value for the significance, or additional values for the power or effect size of the test. Results can be the following type :

• Significance, power, effect size, statistical metrics (such as t, z, T^2 etc) per population
• Classification of the data, either probabilistic or hard, per object
• Intermediate statistical data from the tests, such as classification planes, linear multivariate models etc. Their definition is fully left to the developer of a statistical test.

Local Tests

Local tests return a map/vector/array of values, each representing a single location of region. For each location or region there are the same possible results as in the global tests.

• Surface data results: Stored and visualized as AttributedMeshSpatialObject (Object level description). Default pluggable colormaps for p-values, z-scores, t, T^2 scores etc.

IO contents

• ID / name of test
• Global results are stored directly in datastructure(?)
• Local results are stored in separate files that are referenced in datastrusture
  • e.g. Image data as 3D float data in any ITK readable format
  • e.g. Surface data as SurfaceSpatialObject based class (MetaIO format)