alexholcombe / mixrsvp

Mixture model temporal errors in RSVP data

mixRSVP

mixRSVP mixture models temporal errors in reports of an item from an array or a stream, presently using a mixture model of uniform and Gaussian distributions. Here is an explanation of what the hell that means.

Installation

install_packages('devtools')  #if you don't have devtools already
library('devtools') 
devtools::install_github('alexholcombe/mixRSVP',build_vignettes=TRUE)

Example

Fit some of one of the provided datasets and plot the histogram of serial position errors together with the fitted curve.

 df <-  subset(P2E2pilot, subject=="CB" & target==1 & condition==1)
 plot_hist_with_fit(df, -11, 11, df$targetSP, 16, TRUE, TRUE, TRUE)

Help

Type ?mixRSVP or ?<FUNCTIONNAME> and look at the vignettes that show examples of what you can do.

vignette(package='mixRSVP')

History

Patrick Goodbourn programmed mixture modeling of RSVP serial position errors in MATLAB. Certifiedwaif did initial port of this to R, largely using automated code translation, and then Alex functionfied and improved things before making it into this package.

mixrsvp's Issues

Vignettes stopped working

The vignettes page is still created but the actual vignettes and the links to them from that page aren't. Maybe it happened when I was creating data-format to explain the data format, which is now temporarily stashed in the dev/ directory.

Fix up tests

"hybrid observer" test in test-model-comparison.R changed p-values from 0.2 and I dont know why.

Also test-fit_model.R:5: skip: Check null warning no longer occurs
is an Empty test

Vignettes take too long to build

This is quite annoying because installing source from github, it seems you have to build the vignettes from scratch with

devtools::install_github('alexholcombe/mixRSVP', build = TRUE, build_opts = c("--no-resave-data", "--no-manual"),force=TRUE)

I assume the time-consuming one is analyze-experiment so try to shorten that one.