Taba

Taba calculates robust correlation measures and performs related statistical analyses for continuous and ordinal biological data, enabling detection of associations resilient to outliers and non-normal distributions.

Key Features:

• Robust correlation estimators: Implements Taba (T), TabWil (TW), and TabWil rank (TWR) to measure linear and monotonic associations with robustness to outliers.
• Comparison with classical and robust methods: Evaluates performance against Pearson, Spearman, Quadrant, Median, and Minimum Covariance Determinant (MCD) methods using simulation-based metrics.
• Simulation-based performance metrics: Assesses estimators using root mean square error (RMSE) and bias derived from simulation studies.
• Multiple correlations and p-values: Computes multiple correlations and corresponding p-values simultaneously across variables.
• Partial and generalized partial correlations: Calculates partial, semipartial, and generalized partial correlations by adjusting variables via linear, logistic, or Poisson regression.
• Statistical testing and gene analysis: Performs statistical tests to identify significant associations and has been applied to gene analysis identifying TBL2 as associated with Williams Syndrome.

Scientific Applications:

• Genetics and bioinformatics: Robust correlation measures for analysis of continuous or ordinal genetic and bioinformatics datasets.
• Outlier- and non-normal-aware association analysis: Detection of associations in datasets with outliers or non-normal distributions where traditional correlations may fail.
• Gene association and diagnostic research: Identification of disease-associated genes, exemplified by the association of TBL2 with Williams Syndrome.

Methodology:

Implemented as an R package; uses simulation studies to compare correlation estimators via RMSE and bias across distribution types including bivariate Log-Normal and Weibull; computes partial, semipartial, and generalized partial correlations by fitting linear, logistic, or Poisson regression models; computes multiple correlations, p-values, and performs statistical tests.