fitdistr.core — generateme/fitdistr 1.0.2

Distribution fitting using MLE, MGE and QME methods.

bootstrap clj

(bootstrap method distribution data)

(bootstrap method
           distribution
           data
           {:keys [size samples ci-type all-params? assert?]
            :or {samples 100
                 size (m/constrain (* 0.1 (count data)) 100 5000)
                 ci-type :mad-median
                 all-params? false
                 assert? true}
            :as all})

Bootstrapped version of fitting.

Parameters are the same as fit

Additional parameters:

• :samples - number of bootstrapped sequences (default: 100)
• :size - number of samples in each sequence (default: 10% of data, minimum 100, maximum 5000 samples)
• :ci-type - confidence interval type (default: :mad-median)
• :all-params? - fitted parameters for each sequence (default: false)

fit clj

(fit method distribution data)

(fit method distribution data {:keys [assert?] :or {assert? true} :as all})

Fit distribution using given method

• :mle - log likelihood
• :ad - Anderson-Darling
• :adr, :adl, :ad2r, :ad2l and :ad2 - Anderson-Darling variants
• :ks - Kolmogorov-Smirnov
• :cvm - Cramer-von-Mises
• :qme - quantile matching estimation.
• :mme - method of moments (modified)
• :mps - maximum product of spacing estimation

For QME additional parameters can be provided:

• quantiles - list of quantiles used to match or number of uniformly distributed (default: 50).
• strategy - quantile calculation strategy (default: :legacy).

More about strategies:

• https://generateme.github.io/fastmath/fastmath.stats.html#var-estimation-strategies-list
• http://commons.apache.org/proper/commons-math/javadocs/api-3.6.1/org/apache/commons/math3/stat/descriptive/rank/Percentile.EstimationType.html

Other parameters and distribution names see README

infer clj

(infer distribution data)

(infer distribution data {:keys [assert?] :or {assert? true} :as params})

Infer parameters computationally