opal-analytics-stan

Installation

pip install git+https://github.com/Bedrock-py/opal-analytics-stan.git

Parameters Spec for Bedrock

self.parameters_spec = [
    { "name" : "Formula", "attrname" : "formula", "value" : "decision0d1c ~ condition", "type" : "input" },
    { "name" : "GLM family", "attrname" : "family", "value" : 'logit', "type" : "input" },
    { "name" : "chains", "attrname" : "chains" , "value" : "3", "type" : "input"},
    { "name" : "iter", "attrname" : "iter" , "value" : "3000", "type" : "input"},
    { "name" : "prior", "attrname" : "prior" , "value" : "", "type" : "input"},
    { "name" : "prior_intercept", "attrname" : "prior_intercept" , "value" : "", "type" : "input"}
]

• formula A R-style formula for regression given as a string
• family binomial(link = "logit") is the only family supported currently, future updates will add more families
• chains number of chains to run
• iter number of iterations
• prior expected format: student_t(-0.186, 0.036, df = 7)
• prior intercept expected fromat: student_t(0.662, 0.196, df = 7)

parameters derived from https://github.com/gallup/NGS2/blob/master/NGS2_WITNESS_Cycle1_bayesian_exp1.R

Requires a Matrix with the following files

• matrix.csv The full matrix with both endogenous and exogenous variables
• features.txt A list of column names for the matrix (one name per row)

Outputs the following files

matrix.csv data frame of summary() function on bayesian model prior_summary.txt Text summarizing prior information for model