eddjberry / mixturer

R package to estimate the mixture model described in Bays 2009 for analogue report data

mixturer

Overview

Translation of Paul Bays' Matlab functions for modelling precision data into R. The code is also as close as possible to the Matlab code 'under the hood'. The model implemented is described in:

The precision of visual working memory is set by allocation of a shared resource. Bays PM, Catalao RFG & Husain M. Journal of Vision 9(10): 7, 1-11 (2009).

Installation

remotes::install_github('eddjberry/mixturer')

Example

library(mixturer)

set.seed(1295)

response <- randomvonmises(1000, 0, 1)
target <- response + rnorm(1000, 0, 0.01)

df_test_von_mises <- data.frame(response, target)

df_test_von_mises$response_rad <- wrap(df_test_von_mises$response/90*pi)
df_test_von_mises$target_rad <- wrap(df_test_von_mises$target/90*pi)

(ests1 = bays_2009_fit(X = df_test_von_mises$response_rad, Tg = df_test_von_mises$target_rad, return.ll= F))

##   K   Pt Pn   Pu
## 1 1 0.99  0 0.01

Comparison to Matlab functions

The function names have been changes to have the prefix bays_2009_ to hopefully make it obvious that model is being used. The table below shows the Matlab function names as described in Bays' guide, and the equivalent function in this package.

Matlab function	R function
JV10_error()	bays_2009_error()
JV10_fit()	bays_2009_fit()

Citation

Bays PM, Catalao RFG & Husain M. Journal of Vision 9(10): 7, 1-11 (2009).

Berry E (2019). mixturer: Mixture Models for Analogue Response Data. R package version 0.1.0.

Contributing

I did the orignal translation used for this package back in 2016. It could certainly been improved upon. In addition, there are a number of other models for continuous response data that have been proposed. If someone wanted to, the other models that Paul Bays has proposed could be translation into R and added to this package.

mixturer's Issues

Clean up internal code

This code was initially written in 2016. While it all works it could be cleaned up and made more readable

Better examples

Put some better examples in the README

Fix bays_2009_fit() length warning

Currently calling bays_2009_fit() throws a warning about

number of rows of result is not a multiple of vector length

multiple times as it's looping through doing stuff.

Issue with sd2k function

Line 17 of sd2k function needs parentheses on the denominator. Currently, we get an error:

sd2k(0.3659424)
[1] 0.5295116

This value should return a K of 8. Correcting line 17 to the following fixes the issue:

K = 1 / (R^3 - 4 * R^2 + 3 * R)

sd2k(0.3659424)
[1] 7.995975

Fix tests inside function

Some of the tests inside the functions don't actually work and need fixing