Class DiffExAnalyzerUtilsSampleSelectionTest

java.lang.Object
ubic.gemma.core.analysis.expression.diff.DiffExAnalyzerUtilsSampleSelectionTest
public class DiffExAnalyzerUtilsSampleSelectionTest extends Object
Samples marked DE_Exclude and samples whose assays are flagged as outliers must leave the analysis before anything is decided from the sample set. Everything downstream — baseline selection, dropIncompleteFactors, the analysis-type choice and the degrees-of-freedom guard — reads that set.
Author:
paul

  • Constructor Details

    • DiffExAnalyzerUtilsSampleSelectionTest

      public DiffExAnalyzerUtilsSampleSelectionTest()

  • Method Details

    • setUp

      @BeforeEach public void setUp()

    • testAnOrdinarySampleIsAnalyzed

      @Test public void testAnOrdinarySampleIsAnalyzed()

    • testDeExcludedSampleIsNotAnalyzed

      @Test public void testDeExcludedSampleIsNotAnalyzed()

    • testOutlierSampleIsNotAnalyzed

      @Test public void testOutlierSampleIsNotAnalyzed()

    • testMatrixIsUntouchedWhenThereIsNothingToDrop

      @Test public void testMatrixIsUntouchedWhenThereIsNothingToDrop()

    • testDropsBothDeExcludedAndOutlierSamples

      @Test public void testDropsBothDeExcludedAndOutlierSamples()

    • testPopulateFactorValuesSeesOnlyTheSamplesGiven

      @Test public void testPopulateFactorValuesSeesOnlyTheSamplesGiven()
      Level counting must use the samples that will be modelled. A factor whose second level is carried only by an excluded sample reads as variable over the raw set, survives the subset fix, and becomes a constant column in the model.

    • testBlockCompleteSeesOnlyTheSamplesGiven

      @Test public void testBlockCompleteSeesOnlyTheSamplesGiven()
      A design that is block-complete over every sample need not be over the modelled ones. The interaction term is chosen here and fitted later on the smaller set, so this decision has to see the same samples the fit will.

    • testCheckValidForLmSeesOnlyTheAnalyzedSamples

      @Test public void testCheckValidForLmSeesOnlyTheAnalyzedSamples()
      The point of dropping early: a factor whose second level exists only on samples that will not be modelled must not read as analyzable.