tests/testthat/test-fit.and.evaluate.R

In OmicsQC: Nominating Quality Control Outliers in Genomic Profiling Studies

# Here we want to test
# 1. Input checking
# 2. Whether simulated data is fitted with the correct distribution

# Simulating scores
set.seed(314);
scores <- data.frame(
  'A' = rnorm(100, mean = 2, sd = 3),
  'B' = rnorm(100, mean = 3, sd = 5),
  'C' = rnorm(100, mean = -5, sd = 7)
);
rownames(scores) <- paste0('Patient_', 1:100);

# Signs
signs <- data.frame(
  'Metric' = c('A', 'B', 'C'),
  'Sign' = c('pos', 'neg', 'pos')
);

# Data Processing
zscores <- zscores.from.metrics(scores);
corrected.zscores <- correct.zscore.signs(
  zscores = zscores,
  signs.data = signs,
  metric.col.name = 'Metric',
  signs.col.name = 'Sign'
);
accumulated.zscores <- accumulate.zscores(corrected.zscores);

# Tests
test_that('fit.and.evaluate', {

  # Expected output
  # weibull, gamma, exp, lnorm can only accept positive values
  # todo: fix this
  res <- fit.and.evaluate(
    quality.scores = accumulated.zscores,
    distributions = c('norm', 'cauchy', 'logis'),
    trim.factor = 0.05
  );
  expect_equal(
    colnames(res),
    c('distribution', 'KS.rejected', 'BIC.value')
  );
  expect_true(
    is.character(res$distribution)
  );
  expect_true(
    setequal(res$distribution, c('norm', 'cauchy', 'logis'))
  );
  expect_true(
    is.logical(res$KS.rejected)
  );
  expect_true(
    is.numeric(res$BIC.value)
  );

  # Bad trim factor
  expect_error(
    fit.and.evaluate(
      quality.scores = accumulated.zscores,
      distributions = c('weibull', 'norm', 'gamma', 'exp', 'lnorm', 'cauchy', 'logis'),
      trim.factor = 0.9
    )
  );

  # No data left after trimming
  expect_error(
    fit.and.evaluate(
      quality.scores = accumulated.zscores,
      distributions = c('weibull', 'norm', 'gamma', 'exp', 'lnorm', 'cauchy', 'logis'),
      trim.factor = 0.5
    )
  );

  # Bad input
  bad.input <- accumulated.zscores;
  bad.input$Sum <- sample(LETTERS, size = nrow(bad.input), replace = T);
  expect_error(
    fit.and.evaluate(
      quality.scores = bad.input,
      distributions = c('norm'),
      trim.factor = 0.05
    )
  );

  bad.input <- accumulated.zscores;
  colnames(bad.input)[colnames(bad.input) == 'Sum'] <- 'XX';
  expect_error(
    fit.and.evaluate(
      quality.scores = bad.input,
      distributions = c('norm'),
      trim.factor = 0.05
    )
  );

  # Incorrect distributions
  expect_error(
    fit.and.evaluate(
      quality.scores = accumulated.zscores,
      distributions = c('Norm'),
      trim.factor = 0.05
    )
  );


});