think-bayes

think-bayes

An algorithm collection of probability and statistics for browser and Node.js environment.

In progress...

适用于 浏览器 和 Node.js 环境的概率统计算法集（非正式版本，功能亟待完善，努力 coding 中...）

Install

yarn add think-bayes # OR npm i --save think-bayes

Quickstart

Let us resolve the cookie problem by using the class Suite:

const { Suite } = require('think-bayes');

class Cookie extends Suite {
  mixes = {
    Bowl1: {
      vanilla: 0.75,
      chocolate: 0.25,
    },
    Bowl2: {
      vanilla: 0.5,
      chocolate: 0.5,
    },
  };

  likelihood(data, hypo) {
    const mix = this.mixes[hypo];
    const like = mix[data];
    return like;
  }
}

const hypos = ['Bowl1', 'Bowl2'];
const pmf = new Cookie(hypos);
pmf.update('vanilla');

const result = pmf.render();
console.log(result); // [ [ 'Bowl1', 0.6 ], [ 'Bowl2', 0.4 ] ]

// You can also print the result as a table
pmf.print();
// | Value | Prob |
// |-------|------|
// | Bowl1 | 0.6  |
// | Bowl2 | 0.4  |

In addition, here are some simple demos you can refer directly to resolve some classic problems of probability and statistics.

Algorithm Classes

This library provides some ES Classes following for calculations related to probability and statistics.

These classes can be imported by the same way following:

import { Pmf, Cdf, Pdf, Suite } from 'think-bayes';

DictWrapper(values, name)

An base class for generation an object contains a dictionary.

@Params:

param	type	description
values	string	array
name	string	sequence of values

@Methods:

.initSequence(values)

Initializes with a sequence of equally-likely values.

@Params:

param	type	description
values	array	sequence of values

.initMapping(values)

Initializes with a map from value to probability.

@Params:

param	type	description
values	map	map from value to probability

.initPmf(values)

Initializes with a Pmf.

@Params:

param	type	description
values	pmf	Pmf object

.initFailure(values)

Throw an error.

.values()

Gets an unsorted sequence of values.

Note: One source of confusion is that the keys of this

dictionary are the values of the Hist/Pmf, and the

values of the dictionary are frequencies/probabilities.

.items()

Gets an unsorted sequence of (value, freq/prob) pairs.

.set(value, prob)

Sets the freq/prob associated with the value x.

@Params:

param	type	description
value	any	number value or case name
prob	number	number freq or prob

.incr(x, term = 1)

Increments the freq/prob associated with the value x.

@Params:

param	type	description
x	any	number value or case name
term	number	how much to increment by

.mult(x, factor = 1)

Scales the freq/prob associated with the value x.

@Params:

param	type	description
x	any	number value or case name
factor	number	how much to multiply by

.remove(value)

Removes a value.

Throws an exception if the value is not there.

@Params:

param	type	description
value	any	value to remove

.total()

Returns the total of the frequencies/probabilities in the map.

.maxLike()

Returns the largest frequency/probability in the map.

.copy(name)

Returns a copy.

Make a shallow copy of d. If you want a deep copy of d,

use one method to deep clone the whole object.

@Params:

param	type	description
name	string	string name for the new Hist

@Returns: new object

.scale(factor)

Multiplies the values by a factor.

@Params:

param	type	description
factor	number	what to multiply by

@Returns: new object

.log(m)

Log transforms the probabilities.

Removes values with probability 0.

Normalizes so that the largest logprob is 0.

@Params:

param	type	description
m	number	how much to shift the ps before exponentiating

.exp(m)

Exponentiates the probabilities.

If m is un-exist, normalizes so that the largest prob is 1.

@Params:

param	type	description
m	number	how much to shift the ps before exponentiating

.getDict()

Gets the dictionary.

.setDict(d)

Sets the dictionary.

@Params:

param	type	description
d	map	object

.render()

Generates a sequence of points suitable for plotting.

@Returns: array of [sorted value sequence, freq/prob sequence]

.print()

Prints the values and freqs/probs in ascending order.

@Params:

param	type	description
indent

Pmf(values, name)

Represents a probability mass function.

Values can be any hashable type; probabilities are floating-point.

Pmfs are not necessarily normalized.

@Params:

param	type	description
values	string	array
name	string	sequence of values

@Methods:

Important: This class inherits from DictWrapper, so you can use all methods of the parent class.

.prob(x, probDefault = 0)

Gets the probability associated with the value x.

@Params:

param	type	description
x	any	number value
probDefault	number	value to return if the key is not there

@Returns: probability

.probs(xs)

Gets probabilities for a sequence of values.

@Params:

param	type	description
xs	array	a sequence of values

@Returns: array of probabilities

.makeCdf(name)

Makes a cdf.

@Params:

param	type	description
name	string	the name for new cdf

@Returns: one new cdf

.probGreater(x)

Calculate the probability while the value is greater than x.

@Params:

param	type	description
x	number

@Returns: probability

.probLess(x)

Calculate the probability while the value is less than x.

@Params:

param	type	description
x	number

@Returns: probability

.normalize(fraction = 1.0)

Normalizes this PMF so the sum of all probs is fraction.

@Params:

param	type	description
fraction	number	what the total should be after normalization

@Returns: the total probability before normalizing

.random()

Chooses a random element from this PMF.

@Returns: float value from the pmf

.mean()

Computes the mean of a PMF.

@Returns: float mean

.var(miu)

Computes the variance of a PMF.

@Params:

param	type	description
miu	number	the point around which the variance is computed; if omitted, computes the mean

@Returns: float variance

.maximumLikelihood()

Returns the value with the highest probability.

@Returns: float probability

.credibleInterval(percentage = 90)

Computes the central credible interval.

If percentage=90, computes the 90% CI.

@Params:

param	type	description
percentage	number	float between 0 and 100

@Returns: sequence of two floats, low and high

.add(other)

Computes the Pmf of the sum of values drawn from self and other.

@Params:

param	type	description
other	number	pmf

@Returns: new pmf

.addPmf(other)

Computes the Pmf of the sum of values drawn from self and other.

@Params:

param	type	description
other	pmf	another pmf

@Returns: new pmf

.addConstant(other)

Computes the Pmf of the sum a constant and values from self.

@Params:

param	type	description
other	number	a number

@Returns: new pmf

.sub(other)

Computes the Pmf of the diff of values drawn from self and other.

@Params:

param	type	description
other	pmf	another pmf

@Returns: new pmf

.max(k)

Computes the CDF of the maximum of k selections from this dist.

@Params:

param	type	description
k	number	int

@Returns: new cdf

Cdf(xs, ps, name)

Represents a cumulative distribution function.

@Params:

param	type	description
xs	array	sequence of values
ps	array	sequence of probabilities
name	string	string used as a graph label

@Methods:

Important: This class inherits from DictWrapper, so you can use all methods of the parent class.

.copy(name)

Represents a cumulative distribution function.

@Params:

param	type	description
name	string	string name for the new cdf

@Returns: new cdf

.makePmf(name)

Makes a Pmf.

@Params:

param	type	description
name	string	string name for the new pmf

@Returns: new pmf

.values()

Returns a sorted list of values.

@Returns: array of values

.items()

Returns a sorted sequence of [value, probability] pairs.

@Returns: array of [value, probability] pairs

.append(x, p)

Add an (x, p) pair to the end of this CDF.

Note: this us normally used to build a CDF from scratch, not

to modify existing CDFs. It is up to the caller to make sure

that the result is a legal CDF.

@Params:

param	type	description
x	any	number value or case name
p	number	number freq or prob

.shift(term)

Adds a term to the xs.

@Params:

param	type	description
term	number	how much to add

@Returns: another cdf

.scale(factor)

Multiplies the xs by a factor.

@Params:

param	type	description
factor		what to multiply by

@Returns: another cdf

.prob(x)

Returns CDF(x), the probability that corresponds to value x.

@Params:

param	type	description
x	number	number

@Returns: float probability

.value(p)

Returns InverseCDF(p), the value that corresponds to probability p.

@Params:

param	type	description
p	number	number in the range [0, 1]

@Returns: number value

.percentile(p)

Returns the value that corresponds to percentile p.

@Params:

param	type	description
p	number	number in the range [0, 100]

@Returns: number value

.random()

Chooses a random value from this distribution.

@Returns: number value

.sample(n)

Generates a random sample from this distribution.

@Params:

param	type	description
n	number	int length of the sample

@Returns: array of random values

.mean()

Computes the mean of a CDF.

@Returns: float mean

.credibleInterval(percentage = 90)

Computes the central credible interval.

If percentage=90, computes the 90% CI.

@Params:

param	type	description
percentage	number	float between 0 and 100

@Returns: sequence of two floats, low and high

.render()

Generates a sequence of points suitable for plotting.

An empirical CDF is a step function; linear interpolation can be misleading.

@Returns: array of points

.max(k)

Computes the CDF of the maximum of k selections from this dist.

@Params:

param	type	description
k	number	int

@Returns: new Cdf

Pdf()

Represents a probability density function (PDF).

@Methods:

.density(x)

Evaluates this pdf at x.

This method needs implement by children class, if not there is an UnimplementedMethodException would be throw when the method is called

@Params:

param	type	description
x	number	number

@Returns: float probability density

.makePmf(xs, name)

Makes a discrete version of this pdf, evaluated at xs.

@Params:

param	type	description
xs	string	array

@Returns: new pmf

Suite(values, name)

Represents a suite of hypotheses and their probabilities.

@Params:

param	type	description
values	string	array
name	string	sequence of values

@Methods:

Important: This class inherits from Pmf, so you can use all methods of the parent class.

.update(data)

Updates each hypothesis based on the data.

@Params:

param	type	description
data	any	any representation of the data

@Returns: the normalizing constant

.logUpdate(data)

Updates a suite of hypotheses based on new data.

Modifies the suite directly; if you want to keep the original, make a copy.

Note: unlike Update, LogUpdate does not normalize.

@Params:

param	type	description
any	any	representation of the data

.updateSet(dataset)

Updates each hypothesis based on the dataset.

This is more efficient than calling Update repeatedly because

it waits until the end to Normalize.

Modifies the suite directly; if you want to keep the original, make a copy.

@Params:

param	type	description
dataset	array	set

@Returns: the normalizing constant

.logUpdateSet(dataset)

Updates each hypothesis based on the dataset.

Modifies the suite directly; if you want to keep the original, make a copy.

@Params:

param	type	description
dataset	array	set

.likelihood(data, hypo)

Computes the likelihood of the data under the hypothesis.

This method needs implement by children class

if not there is an UnimplementedMethodException would be throw

@Params:

param	type	description
data	any	some representation of the data
hypo	any	some representation of the hypothesis

@Returns: likelihood

.logLikelihood(data, hypo)

Computes the log likelihood of the data under the hypothesis.

This method needs implement by children class

if not there is an UnimplementedMethodException would be throw

@Params:

param	type	description
data	any	some representation of the data
hypo	any	some representation of the hypothesis

@Returns: likelihood

.makeOdds()

Transforms from probabilities to odds.

Values with prob=0 are removed.

.makeProbs()

Transforms from odds to probabilities.

Hist(values, name)

Represents a histogram, which is a map from values to frequencies.

Values can be any hashable type; frequencies are integer counters.

@Params:

param	type	description
values	string	array
name	string	sequence of values

@Methods:

Important: This class inherits from DictWrapper, so you can use all methods of the parent class.

.freq(x)

Gets the frequency associated with the value x.

@Params:

param	type	description
x	any	number value

@Returns: int frequency

.freqs(xs)

Gets frequencies for a sequence of values.

.isSubset(other)

Checks whether the values in this histogram are a subset of

the values in the given histogram.

.subtract(other)

Subtracts the values in the given histogram from this histogram.

Interpolater(xs, ys)

Represents a mapping between sorted sequences; performs linear interp.

@Params:

param	type	description
xs	array	sorted list
ys	array	sorted list

@Methods:

.lookup(x)

Looks up x and returns the corresponding value of y.

.reverse(x)

Looks up y and returns the corresponding value of x.

Joint(values, name)

Represents a joint distribution.

The values are sequences (usually tuples)

@Params:

param	type	description
values	string	array
name	string	sequence of values

@Methods:

Important: This class inherits from Pmf, so you can use all methods of the parent class.

.marginal(i, name)

Gets the marginal distribution of the indicated variable.

@Params:

param	type	description
i	number	index of the variable we want

@Returns: Pmf

.conditional(i, j, val, name)

Gets the conditional distribution of the indicated variable.

Distribution of vs[i], conditioned on vs[j] = val.

@Params:

param	type	description
i	number	index of the variable we want
j	number	which variable is conditioned on
val		the value the jth variable has to have

@Returns: Pmf

.maxLikeInterval(percentage = 90)

Returns the maximum-likelihood credible interval.

If percentage=90, computes a 90% CI containing the values

with the highest likelihoods.

@Params:

param	type	description
percentage	number	float between 0 and 100

@Returns: list of values from the suite

GaussianPdf()

Represents the PDF of a Gaussian distribution.

@Methods:

Important: This class inherits from Pdf, so you can use all methods of the parent class.

.constructor(mu, sigma)

Constructs a Gaussian Pdf with given mu and sigma.

@Params:

param	type	description
mu	number	mean
sigma	number	standard deviation

.density(x)

Evaluates this Pdf at x.

@Returns: float probability density

GaussianKde()

TODO: implemente this class.

EstimatedPdf(sample)

Represents a PDF estimated by KDE.

Estimates the density function based on a sample.

@Params:

param	type	description
sample	array	sequence of data

@Methods:

Important: This class inherits from Pdf, so you can use all methods of the parent class.

.density(x)

Evaluates this Pdf at x.

@Returns: float probability density

Utility Functions

This library provides some Utility Functions following for calculations related to probability and statistics.

These functions can be imported by the same way following:

import { odds, probability, percentile } from 'think-bayes/helpers';

odds(p)

Computes odds for a given probability.

Example: p=0.75 means 75 for and 25 against, or 3:1 odds in favor.

Note: when p=1, the formula for odds divides by zero, which is

normally undefined. But I think it is reasonable to define Odds(1)

to be infinity, so that's what this function does.

@Params:

param	type	description
p	number	float 0~1

@Returns: float odds

probability(o)

Computes the probability corresponding to given odds.

Example: o=2 means 2:1 odds in favor, or 2/3 probability

@Params:

param	type	description
o	number	float odds, strictly positive

@Returns: float probability

probability2(yes, no)

Computes the probability corresponding to given odds.

Example: yes=2, no=1 means 2:1 odds in favor, or 2/3 probability.

@Params:

param	type	description
yes	number	int or float odds in favor
no	number	int or float odds in favor

percentile(pmf, percentage)

Computes a percentile of a given Pmf.

@Params:

param	type	description
pmf	pmf
percentage	number	float 0-100

credibleInterval(pmf, percentage = 90)

Computes a credible interval for a given distribution.

If percentage=90, computes the 90% CI.

@Params:

param	type	description
pmf	pmf	Pmf object representing a posterior distribution
percentage	number	float between 0 and 100

@Returns: sequence of two floats, low and high

pmfProbLess(pmf1, pmf2)

Probability that a value from pmf1 is less than a value from pmf2.

@Params:

param	type	description
pmf1	pmf	Pmf object
pmf2	pmf	Pmf object

@Returns: float probability

pmfProbGreater(pmf1, pmf2)

Probability that a value from pmf1 is greater than a value from pmf2.

@Params:

param	type	description
pmf1	pmf	Pmf object
pmf2	pmf	Pmf object

@Returns: float probability

pmfProbEqual(pmf1, pmf2)

Probability that a value from pmf1 equals a value from pmf2.

@Params:

param	type	description
pmf1	pmf	Pmf object
pmf2	pmf	Pmf object

@Returns: float probability

randomSum(dists)

Chooses a random value from each dist and returns the sum.

@Params:

param	type	description
dists	array	sequence of Pmf or Cdf objects

@Returns: numerical sum

sampleSum(dists, n)

Draws a sample of sums from a list of distributions.

@Params:

param	type	description
dists	array	sequence of Pmf or Cdf objects
n	number	sample size

@Returns: new Pmf of sums

evalGaussianPdf(x, mu, sigma)

Computes the unnormalized PDF of the normal distribution.

@Params:

param	type	description
x	number	value
mu	number	mean
sigma	number	standard deviation

@Returns: float probability density

makeGaussianPdf(mu, sigma, numSigmas, n = 201)

Makes a PMF discrete approx to a Gaussian distribution.

@Params:

param	type	description
mu	number	float mean
sigma	number	float standard deviation
numSigmas	number	how many sigmas to extend in each direction
n	number	number of values in the Pmf

@Returns: normalized Pmf

evalBinomialPmf(k, n, p)

Evaluates the binomial pmf.

@Returns: the probabily of k successes in n trials with probability p.

evalPoissonPmf(k, lam)

Computes the Poisson PMF.

@Params:

param	type	description
k	number	number of events
lam	number	parameter lambda in events per unit time

@Returns: float probability

Q&A

How to reduce the precision loss caused by the calculation of float point number in javascript?

This library use decimal.js to handle the problem what calculation of float point number, in the same way, you can use it in this library:

import { Decimal } from 'think-bayes';

Decimal.add(0.1, 0.2).toNumber() === 0.3; // true