In simonecollier/lizardHMM: Fitting Lizard Movement Data with HMMs

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)
library(tidyverse)
library(lme4)
library(grid)
library(gridExtra)
library(dplyr)
library(forecast)

lizardHMM

The goal of lizardHMM is to fit lizard movement time series data, composed of step-lengths per second, with hidden Markov models and investigate the quality of fit that arises. This package can work with other time series data including simulated data from the package itself.

Installation

You can install the development version from GitHub with:

# install.packages("devtools")
devtools::install_github("simonecollier/lizardHMM")

Key components

• Simulate data from an HMM

• norm_generate_sample() produces data from an n-state HMM with the desired normal state dependent distributions and transition probabilities.

• Fit data with an HMM

• norm_fit_hmm() and gam_fit_hmm() both work to fit an n-state HMM with
  normal/gamma state dependent distributions to the data given.
• gam0_fit_hmm() fits the data with an n-state HMM with gamma state dependent distributions and includes a point mass on zero for the state with the smallest mean.

• norm_viterbi(), gam_viterbi(), and gam0_viterbi() use global decoding to find the most likely sequence of states that could have generated the data according the HMM that was fit.

• Check fit

• norm_ci(), gam_ci(), and gam0_ci() produce confidence intervals for each of the fitted parameters.

• norm_forecast_psr(), gam_forecast_psr(), and gam0_forecast_psr() compute the normal forecast pseudo-residuals for the data fitted with the HMM.

• Visualizations

• timeseries_plot() plots the time series data with colors corresponding to the states decoded by the viterbi algorithm.
• norm_hist_ci(), gam_hist_ci(), and gam0_hist_ci() plot the histogram of the data with the fitted state dependent distributions overlayed and their corresponding confidence intervals.
• The functions in psr_plotting.R produce visualizations of the normal forecast pseudo-residuals.

Extra details

All of the distributions types are set up to handle multiple subjects, variables, and covariates for the transition probabilities. The only option is complete pooling of parameters when working with multiple subjects, although there may be updates in the future that include more options. The functions in covariate_analysis.R can be used to investigate the effect of a single covariate on the transition probabilities and stationary distribution.

simonecollier/lizardHMM documentation built on Dec. 23, 2021, 2:24 a.m.