Matrix Tests

A package dedicated to running multiple statistical hypothesis tests on rows and columns of matrices.

Goals

1. Fast execution via vectorization.
2. Convenient and detailed output format.
3. Compatibility with tests implemented in base R.
4. Careful handling of missing values and edge cases.

Examples

1. Bartlett's test on columns

Bartlett's test on every column of iris dataset using Species as groups:

r col_bartlett(iris[,-5], iris$Species) obs.tot obs.groups var.pooled df statistic pvalue Sepal.Length 150 3 0.26500816 2 16.005702 0.0003345076070163084 Sepal.Width 150 3 0.11538776 2 2.091075 0.3515028004158132768 Petal.Length 150 3 0.18518776 2 55.422503 0.0000000000009229038 Petal.Width 150 3 0.04188163 2 39.213114 0.0000000030547839322

2. Welch t-test on rows

Welch t-test performed on each row of 2 large (million row) matrices:

```r X <- matrix(rnorm(10000000), ncol = 10) Y <- matrix(rnorm(10000000), ncol = 10)

rowtwelch(X, Y) # running time: 2.4 seconds ```

Confidence interval computations can be turned-off for further increase in speed:

r row_t_welch(X, Y, conf.level = NA) # running time: 1 second

Available Tests

| Variant | Name | Function | |----------------------------------|---------------------------------------|------------------------------------| | Location tests (1 group) | Single sample Student's t.test | row_t_onesample | | | Single sample Wilcoxon's test | row_wilcoxon_onesample | | Location tests (2 groups) | Equal variance Student's t.test | row_t_equalvar | | | Welch adjusted Student's t.test | row_t_welch | | | Two sample Wilcoxon's test | row_wilcoxon_twosample | | Location tests (paired) | Paired Student's t.test | row_t_paired | | | Paired Wilcoxon's test | row_wilcoxon_paired | | Location tests (2+ groups) | Equal variance oneway anova | row_oneway_equalvar | | | Welch's oneway anova | row_oneway_welch | | | Kruskal-Wallis test | row_kruskalwallis | | | van der Waerden's test | row_waerden | | Scale tests (2 groups) | F variance test | row_f_var | | Scale tests (2+ groups) | Bartlett's test | row_bartlett | | | Fligner-Killeen test | row_flignerkilleen | | | Levene's test | row_levene | | | Brown-Forsythe test | row_brownforsythe | | Association tests | Pearson's correlation test | row_cor_pearson | | Periodicity tests | Cosinor | row_cosinor | | Distribution tests | Kolmogorov-Smirnov test | row_kolmogorovsmirnov_twosample | | Normality tests | Jarque-Bera test | row_jarquebera | | | Anderson-Darling test | row_andersondarling |

Further Information

For more information please refer to the Wiki page:

1. Installation Instructions
2. Design Decisions
3. Speed Benchmarks
4. Bug Fixes and Improvements to Base R

See Also

Literature

Computing thousands of test statistics simultaneously in R, Holger Schwender, Tina Müller.\ Statistical Computing & Graphics. Volume 18, No 1, June 2007.

Packages

CRAN:

1. ttests() in the Rfast package.
2. row.ttest.stat() in the metaMA package.
3. MultiTtest() in the ClassComparison package.
4. bartlettTests() in the heplots package.
5. harmonic.regression() in the HarmonicRegression package.

BioConductor:

1. lmFit() in the limma package.
2. rowttests() in the genefilter package.
3. mt.teststat() in the multtest package.
4. row.T.test() in the HybridMTest package.
5. rowTtest() in the viper package.
6. lmPerGene() in the GSEAlm package.

GitHub:

1. rowWilcoxonTests() in the sanssouci package.
2. matrix.t.test() in the pi0 package.
3. wilcoxauc() in the presto package.