README.md

In hhau/rjmonopoly: Order Selection of Monotonic Polynomials via Reversible Jump

rjmonopoly: Automatic(ish) monotonic polynomial regression

Andrew Manderson

Overview

This package enables users to fit monotonic polynomials of varying degree using a reversible jump sampler.

Installation

This package is not in CRAN right now. Just install using devtools and github

devtools::install_github('hhau/rjmonopoly')

Overview

To fit, simply call the rjmonopoly function:

library(fda)
library(rjmonopoly)
x <- onechild$day
y <- onechild$height  
fit <- rjmonopoly(x, y)

The help file should be reasonably self-explanatory for the other parameters. At the moment there are only rudamentary plotting functions avaliable.

plotDegreePost(fit)
plotFit(fit)

I intend on adding more / some sensible generics, once I figure out what those should be.

NB

This sampler (and code) is currently delicate, and is likely to break, hang, and emit no errors. Sorry about that, it’s on my to do list.

To do:

• A method for finding a good starting value for the variance parameter
• Diagnostics to identify sample chains that haven’t performed appropriately
• Adaptive innovation variances?
• Multiple chains over multiple cores
• A whole lot of these functions should be ported to c++, because some chains need to run for millions of iterations, and no one wants to wait for R to do that.
• ~~Flat prior option.~~

hhau/rjmonopoly documentation built on May 17, 2019, 4:01 p.m.