uhcdensplot: uhcdensplot

In aaarchmiller/uhcplots: Used-Habitat Calibration Plots (UHC Plots)

Description

uhcdensplot creates UHC plots

Usage

uhcdensplot(densdat, densrand, includeAvail = F, densavail = NULL,
  xl = NULL, yl = NULL, includeLegend = T)

Arguments

densdat	The kernel density estimates of observed points in the test data set
densrand	he kernel density estimates for the habitat covariate at the predicted test data points (across M predicted data sets)
includeAvail	An indicator determining whether the distribution of the available locations should be drawn on the plot
densavail	The kernel density estimates of available points in the test data set
xl	The x-axis limits (can be user supplied)
yl	The y-axis limits (can be user supplied)
includeLegend	An indicator determining whether the legend should be displayed.

Details

This function plots the density of the environmental characteristics at the observed locations in the test data set, $f^u(z)$, along with a simulation envelope for $f^U(z)$ created by randomly choosing locations in the test data using the uhcsim or uhcsimstrat function.

Value

A plot with a 95 density estimates for simulated data points and kernel density estimates of the environmental characteristics associated with the observed locations (solid black line) and the available locations (dashed red line).

See Also

Full archive of the data and code necessary to replicate the manuscript at http://doi.org/10.13020/D6T590.

Examples

# Simulate training data for the non-linear example
nonlinear.train <- uhcdatasimulator(nused = 100,
    navail = 10000,
    betas = c(2,-1),
    ntemp = 1000000,
    example = "non-linear")

# Simulate test data for the non-linear example
nonlinear.test <- uhcdatasimulator(nused = 100,
    navail = 10000,
    betas = c(2,-1),
    ntemp = 1000000,
    example = "non-linear")

# Fit GLM with quadratic relationship
train.correct <- glm(y~temp + I(temp^2),
   family = binomial,
   data = nonlinear.train)

# Fit GLM with linear (misspecified) relationship
train.misspec <- glm(y~temp,
   family = binomial,
   data = nonlinear.train)

# Simulate data for quadratic model
xhat.correct <- uhcsim(nsims = 1000,
   nused_test = 100,
   xmat = model.matrix(y~temp + I(temp^2), data = nonlinear.test)[,-1],
   fit_rsf = train.correct,
   z = as.matrix(nonlinear.test[,"temp"]))

# Simulate data for linear (misspecified) model
xhat.misspec <- uhcsim(nsims = 1000,
   nused_test = 100,
   xmat = as.matrix(model.matrix(y~temp, data = nonlinear.test)[,2]),
   fit_rsf = train.misspec,
   z = as.matrix(nonlinear.test[,"temp"]))

# Get density estimates for quadratic model
denshats.correct <- uhcdenscalc(rand_sims = xhat.correct[,,1],
   dat = subset(nonlinear.test, y==1, select="temp"),
   avail = subset(nonlinear.test, y==0, select="temp"))

# Get density estimates for linear (misspecified) model
denshats.misspec <- uhcdenscalc(rand_sims = xhat.misspec[,,1],
   dat = subset(nonlinear.test, y==1, select="temp"),
   avail = subset(nonlinear.test, y==0, select="temp"))

# Create a UHC plot for the quadratic model
uhcdensplot(densdat = denshats.correct$densdat,
   densrand = denshats.correct$densrand,
   includeAvail = TRUE,
   densavail = denshats.correct$densavail,
   includeLegend = TRUE)

# Create a UHC plot for the linear (misspecified) model
uhcdensplot(densdat = denshats.misspec$densdat,
   densrand = denshats.misspec$densrand,
   includeAvail = TRUE,
   densavail = denshats.misspec$densavail,
   includeLegend = TRUE)

aaarchmiller/uhcplots documentation built on May 10, 2019, 2:05 a.m.