statkit

A statistics toolkit for javascript.

Usage

Install using npm:

npm install statkit

Fit a linear regression model using MCMC:

varsk=require("statkit.js");// log-likelihood for the model y ~ N(m*x + b, 1/t)functionlnlike(theta,x,y){varm=theta[0],b=theta[1],t=theta[2];vars=0.0;for(vari=0;i<x.length;i++){varr=y[i]-(m*x[i]+b);s+=r*r*t-Math.log(t);}return-0.5*s;}// uniform log-prior for m, b, tfunctionlnprior(theta){varm=theta[0],b=theta[1],t=theta[2];if(0.0<m&&m<1.0&&0.0<b&&b<10.0&&0.0<t&&t<100.0){return0.0;}return-Infinity;}// posterior log-probability functionfunctionlnpost(theta,x,y){varlp=lnprior(theta);if(!isFinite(lp)){return-Infinity;}returnlp+lnlike(theta,x,y);}varx=[10,8,13,9,11,14,6,4,12,7,5];vary=[8.04,6.95,7.58,8.81,8.33,9.96,7.24,4.26,10.84,4.82,5.68];varres=sk.metropolis(function(theta){returnlnpost(theta,x,y);},[0.5,3.0,1.0],1000000,0.1,50000,100);console.log('acceptance rate:',res.accepted)console.log('posteriors (16/50/84 percentiles):')console.log('m',sk.quantile(res.chain[0],0.16),sk.median(res.chain[0]),sk.quantile(res.chain[0],0.84))console.log('b',sk.quantile(res.chain[1],0.16),sk.median(res.chain[1]),sk.quantile(res.chain[1],0.84))console.log('t',sk.quantile(res.chain[2],0.16),sk.median(res.chain[2]),sk.quantile(res.chain[2],0.84))

Calculate a confidence interval for a correlation using the bootstrap method:

varsk=require("statkit");varlsat=[576,635,558,578,666,580,555,661,651,605,653,575,545,572,594];vargpa=[3.39,3.30,2.81,3.03,3.44,3.07,3.00,3.43,3.36,3.13,3.12,2.74,2.76,2.88,2.96];varcorr=sk.corr(gpa,lsat);varci=sk.bootci(100000,sk.corr,gpa,lsat);console.log("corr = ",corr,"ci = ",ci);

Perform a linear regression on the first data set in Anscombe's quartet:

varsk=require("statkit");varx=[10,8,13,9,11,14,6,4,12,7,5];vary=[8.04,6.95,7.58,8.81,8.33,9.96,7.24,4.26,10.84,4.82,5.68];varA=newArray(x.length*2);for(vari=0;i<x.length;++i){A[2*i]=1;A[2*i+1]=x[i];}varb=sk.lstsq(x.length,2,A,y);console.log("intercept = ",b[0],"slope = ",b[1]);

Functions

• min(a) - Minimum
• max(a) - Maximum
• range(a) - Range
• quantile(a) - Quantile
• median(a) - Median
• iqr(a) - Interquartile range
• mean(a) - Mean
• gmean(a) - Geometric mean
• hmean(a) - Harmonic mean
• var(a) - Variance
• std(a) - Standard deviation
• skew(a) - Skewness
• kurt(a) - Kurtosis
• corr(x, y) - Correlation between x and y
• entropy(p) - Entropy
• kldiv(p, q) - Kullback–Leibler divergence
• shuffle(a) - Shuffle using the Fisher–Yates shuffle
• sample(a) - Sample with replacement
• boot(nboot, bootfun, data...) - Bootstrap the bootfun statistic
• bootci(nboot, bootfun, data...) - Calculate bootstrap confidence intervals using the normal model
• randn() - Draw random sample from the standard normal distribution using the Marsaglia polar method
• normcdf(x) - Normal cumulative distribution function
• norminv(p) - Normal inverse cumulative distribution function
• lufactor(A, n) - Compute pivoted LU decomposition
• lusolve(LU, p, b) - Solve Ax=b given the LU factorization of A
• qrfactor(m, n, A) - Compute QR factorization of A
• qrsolve(m, n, QR, tau, b) - Solve the least squares problem min ||Ax = b|| using QR factorization QR of A
• lstsq(m, n, A, b) - Solve the least squares problemmin ||Ax = b||
• metropolis(lnpost, p, iterations, scale, burn, thin) - Sample from lnpost starting at p using the Metropolis-Hastings algorithm

Credits

(c) 2014 Erik Rigtorp erik@rigtorp.se. MIT License