Title: Goodness of Fit Checks for Binomial N-Mixture Models
Version: 0.1.1
Description: Provides residuals and overdispersion metrics to assess the fit of N-mixture models obtained using the package 'unmarked'. Details on the methods are given in Knape et al. (2017) <doi:10.1101/194340>.
Depends: R (≥ 3.3.0)
License: GPL-3
Encoding: UTF-8
LazyData: true
LinkingTo: Rcpp
Imports: Rcpp, unmarked
RoxygenNote: 7.3.2
URL: https://github.com/jknape/nmixgof
BugReports: https://github.com/jknape/nmixgof/issues
NeedsCompilation: yes
Packaged: 2025-02-12 16:22:03 UTC; jonas
Author: Jonas Knape ORCID iD [aut, cre]
Maintainer: Jonas Knape <jonas.knape@slu.se>
Repository: CRAN
Date/Publication: 2025-02-12 17:50:05 UTC

Goodness of fit checks for binomial N-mixture models

Description

The package contains methods to compute overdispersion metrics, randomized quantile residuals, and graphical diagnostics of model fit for binomial N-mixture models fitted using the unmarked package. Details about the checks are given in Knape et al. (2018).

Author(s)

Maintainer: Jonas Knape jonas.knape@slu.se (ORCID)

References

Knape et al. 2018. Sensitivity of binomial N-mixture models to overdispersion: the importance of assessing model fit. Methods in Ecology and Evolution, 9:2102-2114. doi:10.1111/2041-210X.13062

See Also

Useful links:

Overdispersion metrics for binomial N-mixture models.

Description

Computes various types of overdispersion metrics, based on Pearson residuals, for binomial N-mixture models.

Usage

chat(umFit, type = "marginal")

Arguments

umFit

An object of class unmarkedFit from a model fitted using pcount.

type

The type of metric to compute, one of 'marginal', 'site-sum' or 'observation'.

Value

An estimate of overdispersion relative to the fitted model.

Examples

library(unmarked)
data(mallard)
fm.mallard <- pcount(~ 1 ~ 1, unmarkedFramePCount(y = mallard.y), K=100)
chat(fm.mallard, "m")
chat(fm.mallard, "s")
chat(fm.mallard, "o")

Plot residuals against covariates

Description

A convenience function to plot rq residuals against all untransformed numeric covariates. Site-sum randomized quantile residuals are used for site covariates while marginal residuals are used for observation covariates. The same random residual draws are reused for different covariates.

Usage

residcov(umFit, ...)

Arguments

umFit

An object of class unmarkedFit from a model fitted using pcount.

...

Plot arguments.

Examples

library(unmarked)
umf = unmarkedFramePCount(y = shoveler$y, obsCovs = shoveler$obs, siteCovs = shoveler$site)
fmP = pcount(~scale(date) + scale(reedcover) ~ scale(log(water)) + scale(latitude), 
      data = umf, K = 80)
residcov(fmP)

Plot residuals against fitted values

Description

Plots randomized-quantile residuals for binomial N-mixture models against fitted values.

Usage

residfit(umFit, type = "marginal", ...)

Arguments

umFit

An object from a model fitted using pcount.

type

The type of randomized quantile residual to plot. One of 'marginal', 'site-sum' or 'observation'.

...

Plot arguments.

Examples

library(unmarked)
umf = unmarkedFramePCount(y = shoveler$y, obsCovs = shoveler$obs, siteCovs = shoveler$site)
fmP = pcount(~scale(date) + scale(reedcover) ~ scale(log(water)) + scale(latitude), 
      data = umf, K = 80)
residfit(fmP, "marginal")
residfit(fmP, "site-sum")
residfit(fmP, "observation")

Qq plot of randomized quantile residuals against standard normal quantiles

Description

Qq plot of randomized quantile residuals against standard normal quantiles

Usage

residqq(
  umFit,
  type = "site-sum",
  main = "Residual qq plot",
  plotLine = TRUE,
  ...
)

Arguments

umFit

An object of class unmarkedFit from a model fitted using pcount.

type

The type of randomized quantile residual to plot. One of 'site-sum' or 'observation'.

main

Plot label.

plotLine

If true, the identity line is added to the plot.

...

Further arguments passed to qqnorm.

Value

A list with x and y coordinates of the qq plot, see qqnorm.

Examples

library(unmarked)
umf = unmarkedFramePCount(y = shoveler$y, obsCovs = shoveler$obs, siteCovs = shoveler$site)
fmP = pcount(~scale(date) + scale(reedcover) ~ scale(log(water)) + scale(latitude), 
      data = umf, K = 80)
residqq(fmP, "site-sum")
residqq(fmP, "observation")

Randomized quantile resiudals for binomial N-mixture models.

Description

Computes three types of randomized quantile residuals for binomial N-mixture models.

Usage

rqresiduals(umFit, type = "marginal")

Arguments

umFit

An object of class unmarkedFit from a model fitted using pcount.

type

The type of rq residuals to compute, one of 'marginal', 'site-sum' or 'observation'.

Value

A matrix (if type is 'marginal' or 'site-sum') or vector (for ) con.

Examples

library(unmarked)
umf = unmarkedFramePCount(y = shoveler$y, obsCovs = shoveler$obs, siteCovs = shoveler$site)
fmP = pcount(~scale(date) + scale(reedcover) ~ scale(log(water)) + scale(latitude), 
      data = umf, K = 80)
qqnorm(rqresiduals(fmP, "s"))
qqnorm(rqresiduals(fmP, "o"))
par(mfcol = c(3,4))
invisible(apply(rqresiduals(fmP, "m"), 2, qqnorm))

Northern shoveler data

Description

Repeated count data of Northern shoveler with covariates, formatted for use with the unmarked package.

Usage

shoveler

Format

A list with three elements

y

A matrix with Northern shoveler counts

site

A data frame with site specific covariates

obs

A list containing observation specific covariates

References

Knape et al. (2018) Methods in Ecology and Evolution, 9:2102-2114. doi:10.1111/2041-210X.13062

Examples

library(unmarked)
umf = unmarkedFramePCount(y = shoveler$y, obsCovs = shoveler$obs, siteCovs = shoveler$site)