jimmymandersson / statkit Goto Github PK

Definition of done:

• Expand the information on installing different versions.
• Fix inline code block pod StatKit to say pod 'StatKit'.

Run testing and linting on every pull request aimed towards the master branch.

Sequences should support ranking with the Modified Competition strategy.

https://www.wikiwand.com/en/Ranking

Variance should be factored out from the Distribution protocol and added to a new Univariate Distribution protocol. A covariance requirement for multivariate distributions should also match it.

Definition of done:

• The covariance between two random variables in a collection can be calculated.

Add a protocol to define suitable regression metrics for inferential/predictive statistics.

• Coefficient of determination (R Squared)
• Adjusted R Squared
• MSE
• RMSE
• Mean Absolute Error
• Median Absolute Error
• Max error
• RSS
• pMSE

https://en.wikipedia.org/wiki/Gamma_function

Sequences should support ranking with the Standard Competition strategy.

https://www.wikiwand.com/en/Ranking

Definition of done:

• Supports non-verbose syntax for the most commonly used quantiles (quartile, quintile, decile).
• Supports arbitrary splits.

https://github.com/apple/swift-numerics

https://en.wikipedia.org/wiki/Continuous_uniform_distribution

Extensions to standard library types do not show up in the documentation generated by Swift-DocC.
Even though the added functions make sense in an extension context, they are of no use if no one knows they exist.
They also make sense as free functions and should be rewritten as such.

• Quantile
• Sum
• Range
• Variance
• Standard Deviation
• Rank
• Mean
• Median
• Mode
• Covariance
• Correlation

This protocol should be designed to encompass the commonalities between different distributions and lay the foundation for future functionality in both descriptive and inferential statistics.

Collections with ConvertibleToReal components should be able to calculate Spearman's Rank Correlation Coefficient for two arbitrary components.
https://www.wikiwand.com/en/Spearman%27s_rank_correlation_coefficient

https://en.wikipedia.org/wiki/Poisson_distribution

https://en.wikipedia.org/wiki/Normal_distribution

https://en.wikipedia.org/wiki/Beta_distribution

Simplify complex function for use with real numbers only.
https://www.math.ubc.ca/~cbm/aands/abramowitz_and_stegun.pdf

https://developer.apple.com/documentation/swift/randomnumbergenerator

Definition of done:

• A CI pipeline generates and publishes new documentation on pushes to the main branch.

Possible tools:

• Jazzy
• Swift Doc

According to specifications in:
https://www.wikiwand.com/en/Goodman_and_Kruskal%27s_gamma

https://en.wikipedia.org/wiki/Harmonic_mean

https://en.wikipedia.org/wiki/Beta_function

https://en.wikipedia.org/wiki/F-distribution

https://en.wikipedia.org/wiki/Discrete_uniform_distribution

https://en.wikipedia.org/wiki/Binomial_distribution

Collections with ConvertibleToReal components should be able to calculate Kendall's Rank Correlation Coefficient for two arbitrary components.
https://www.wikiwand.com/en/Kendall_rank_correlation_coefficient

Sequences should support ranking with the Fractional strategy.

https://www.wikiwand.com/en/Ranking

Sequences should support ranking with the Ordinal strategy.

https://www.wikiwand.com/en/Ranking

https://en.wikipedia.org/wiki/Exponential_distribution

https://en.wikipedia.org/wiki/Gamma_distribution

https://en.wikipedia.org/wiki/Geometric_distribution

http://xoshiro.di.unimi.it/xoshiro256starstar.c

Distributions should be able to output logarithmic values if needed.
Use cases include, for example, increasing numerical stability when chaining multiple outputs together.

The arithmetic mean should be available on all collections where the random variable under study conforms to ConvertibleToDouble.

Sequences should support ranking with the Dense strategy.

https://www.wikiwand.com/en/Ranking

Developers should be able to calculate the range of a variable in a data set. The statistic should be made available through the Sequence protocol.

Collections with ConvertibleToReal components should be able to calculate the Pearson Correlation Coefficient for two arbitrary components.

Both the Beta and the Gamma function will need to accept real valued numbers before the Beta and Gamma distributions can be implemented in a satisfactory manner.

https://en.wikipedia.org/wiki/Negative_binomial_distribution

https://en.wikipedia.org/wiki/Student%27s_t-distribution

According to specifications in:
https://www.wikiwand.com/en/Somers%27_D

https://forums.swift.org/t/announcing-swift-docc/52797

https://en.wikipedia.org/wiki/Arcsine_distribution