duanweikang/plan-b

Instructions on how to use the code

contents

This repo has three files:

multinom3.R the code you need to source for use,

simulation_multinom.R simulation code to see how model works,

test_on_data_multi.R code for test model on real data set

prerequisite packages

You need to download partitions package and install it first on R.

Usage

This code is used to implement a nonparametric bayesian regression model for multinomial responses and continuous predictor. The model is based on this paper.

The only function you need to use is manysteps_multinom, note there are several parameters need to be given as input.

y as the response data, should be a matrix of 1 and 0 where rows equals number of categories and columns equals number of observations.

m is the minimum number of each chunk,

k is the number of chunks.

R is the length of MCMC

ct is the number of categories.

Example

> dat = rmultinom(100,1,c(0.3,0.5,0.2))

> foo = manysteps_multinom(dat,m=5,k=10,R=1e3,ct=3)

> foo[[1]]

[1] 0.624

> foo[[2]][,1:4]

[,1] [,2] [,3] [,4]

[1,] 0.47418990 0.47418990 0.47418990 0.47418990

[2,] 0.48429465 0.48429465 0.48429465 0.48429465

[3,] 0.04151545 0.04151545 0.04151545 0.04151545