README.md

In statdivlab/rre: Replicate Richness Estimator

rre

Replicate Richness Estimation

Authors: Alex Paynter and Amy Willis

Introduction

rre is an R package for estimating species richness where multiple frequency count tables have been collected. All methods in this package assume a negative binomial model and explore the use of penalization to leverage the replicate data for improved estimation. rre_sims is a related package containing code for simulations and data analyses.

Installation

rre can be installed from GitHub by running the following code in an R console:

# install.packages("devtools") # run if devtools is not already installed.
devtools::install_github("statdivlab/rre")
library(rre)

Example

In this example we will demonstrate the use of two rre methods using simulated data. The two methods we show are the negative binomial MLE without regularization and a method which uses goodness of fit to tune a penalization term.

Generate data

We generate 2 replicates from a population with 100 species, with the negative binomial parameters alpha = 0.1 and delta = 0.1:

library(rre)
set.seed(12)
list_of_fct <- nb_fct_simulation(ccc = 100, alpha = 0.1, delta = 0.1, r = 2)

MLE

The function unregularized_mle optimizes the negative binomial likelihood with no penalization:

# The function unregularized_mle optimizes the negative binomial 
# likelihood with no penalization:
mle_result <- unregularized_mle(list_of_fct)
# To extract the MLE we extract the 'best' element:
mle_result$best # ccc is the richness estimate, in this case 79

Goodness of fit method

gof_criterion is one of our novel methods, which uses penalization tuned by a goodness of fit metric to estimate richness:

# gof_criterion is one of our novel methods, which uses penalization tuned 
#   by a goodness of fit metric to estimate richness:
gof_result <- gof_criterion(list_of_fct)
# Extract the 'best' element:
gof_result$best # ccc is the richness estimate, in this case 79

Issues

Our penalization tuning methods use grids for both richness and the penalization parameter. A warning message is printed when the estimate is near the upper end of either of these grids, with a recommended course of action.

If you encounter issues please let us know by filing an issue.

statdivlab/rre documentation built on Nov. 5, 2019, 9:20 a.m.