BDA helpers for WebPPL

Data analysis helpers

• observeData({data: data, link: Dist, display: true/false}): e.g., observeData({data: [1,2], link: Binomial({n:2, p:0.5}), display: true}) (display will print out the score to the console). Shorthand for factor(Dist.score(data)) (but works with arrays as well as singleton data points)
• foreach(array, function): like map, but no return (useful for IncrementalMH)
• marginalize(Dist, label): go from a joint distribution to a marginal distribution

Linking functions

• DiscreteGaussian({mu: ..., sigma: ..., lower: ..., upper: ..., binWidth: ...}): a discretized Gaussian disribution with lower and upper bound stipulated [default lower and upper set to 1 and 7]
• OrdinalGaussian({mu: ..., sigma: ..., thresholds: ...}): an ordinalized Gaussian distribution (e.g., for ordinal regression). e.g. OrdinalGaussian({mu: 1, sigma:2.5, thresholds: [1.5, 2.5, 3.5, 4.5, 5.5, 6.5]})
• gaussianCDF({mu: ..., sigma: ..., x: ...}): e.g. gaussianCDF({mu: 3.5, sigma:1, x: 2})

Misc helpers

• exp(x): exponentiation e.g. exp(1.37)
• probability(x, Dist): shorthand for exp(Dist.score(x))

List of .js functions

Accessed via bdaUtils.function, e.g. bdaUtils.isNumeric(x)

• isNumeric(x) where x may or may not be numeric
• parseFloat(x) where x is a string
• fillArray(x, length)
• erf(x) where x is numeric
• closest(x, array): find the closest value to x in array of numbers

Installation

You can install this pacakge using:

npm install --prefix ~/.webppl mhtess/webppl-bda