v2.13.0

New Features

• Matrix columnAddScalar
• Matrix columnSubtract
• Matrix columnSubtractScalar
• Matrix columnDivide

Bug Fixes

• Algebra::quadratic returns empty array when a=0 and b=0 (Previously would divide by zero)
• Trigonometry::unitCircle division by zero error for single-point.
• SampleData\People getShoeSize has a typo and was getShowSize. This has been fixed.
• Fixed calculation in Probability\Distribution\Discrete\Hypergeometric cdf. Implementation changed to use sum of PMFs.
• Statistics\Correlation sampleCovariance, populationCorrelationCoefficient, sampleCorrelationCoefficient, throw BadDataException or OutOfBoundsException rather than division by zero for bad input data.
• Statistics\Descriptive coefficientOfVariation will return NAN if mean is zero. Previously did division by zero.
• Fix some typos in Statistics\Distance hellinger, brayCurtis, and canberra that could have resulted in incorrect results.
• Fix \Statistics\EffectSize cohensQ bounds check [-1, 1] and throw OutOfBoundsException when outside range, and return INF when r == 1 or -1.
• Fix \Statistics\EffectSize boundary edge cases resulting in division by zero for hedgesG, cohensD, and glassDelta.

Improvements

• Refactor the geometric mean implementation to use a numerically stable log-space calculation
• Refactor ChiSquared pdf to avoid raising 0 to negative exponent which is deprecated
• Refactor LogLogistic pdf and cdf to avoid raising 0 to negative exponent which is deprecated
• Improved implementation of NoncentralT median
• Probability\Distribution\Multivariate\Multinomial uses a higher numerical precision to reject probabilities that don't sum to 1.0
• Rewrote Statistics\Distribution fractionalRanking to track original indices and use == for tie detection instead of strval() grouping, avoiding floating-point comparison bugs.
• Statistics\Distribution cumulativeFrequency and cumulativeRelativeFrequency now do a frequency distribution on sorted data, which was the original intent.
• Statistics\Divergence kullbackLeibler and jensenShannon throw an exception if there are non-negative probabilities.
• Statistics\Distribution stemAndLeafPlot throws an exception if there are negative values.
• Statistics\Experiment functions throw exceptions when input outside of range. likelihoodRatioSS returns infinity for rather than division by zero.
• Statistics\Outlier edge cases throw exceptions: Grubbs' having fewer than 3 observations; standard deviation of zero; Support upper/lower as aliases for one sided in grubbsCriticalValue.
• Improved exception message for Functions\Support checkLimits to add invalid endpoint character in the exception message