Batman

Batman

Julia library designed to build statistical models with associated systematic uncertainties and provide inference into particular parameters of interest. Both Bayesian and Frequentist analysis are supported.

abstract type Model

Models hold the parameters of interest and the associated likelihood/cost functions used in optimization. Individual models should be derived from this base object.

NLogLikelihood(pf::Array{NLogPDF})

Given an array of NLogPDF, a complete negative log-likelihood function is contructed which can then be passed to a minimizer.

NLogPDF(f::String, p::Variable...)

Negative log PDF used when contructing the complete likelihood function. The NLogPDF should have a unique name f::String.

Results

Created upon optimizing a model, contains the model and all of its parameters as well as the optimized results found by minimizing the cost function.

@add_dataset(name, df::DataFrame)

Introduce a new dataset into the namespace at read-time.

lognormal(x, μ, σ)

logpoisson(n, x...)

math

optimizemodel!(m::Model, nll::NLogLikelihood; lowerbounds=nothing, upper_bounds=nothing)

Currently NLopt is used for minimizing the objective function. optimizemodel! produces the objective function from the given NLogLikelihood which is then passed to NLopt to be minimized. optimizemodel! returns a set of Results.

@addfunction f

Using @addfunction will write in a new function into the current namespace at read-time.

Examples

addfunction lognormalconstraint(x, μ, σ) = (x-μ)^2/2/σ