ddFit: Fit Distance Distribution Model(s)

Description

Fit generalized linear models (glm) for distance distribution models corresponding to standard forms [xep1, xep01 (gamma), xep2 (Rayleigh), xep02, xep12, xep012, xep123, xep0123 (normal-gamma with x = tau), lognormal, truncated normal, Maxwell Boltzmann, and constant] and supplentary forms [exponential, chi-squared, inverse gamma, and inverse Gaussian].

The glm is converted to a probability distribution by dividing by a normalizing constant, namely the integral of the glm evaluated from 0 to infinity. In some cases (most notably when the leading coefficient of the glm is positive so the fitted curve does not converge to zero as x increases), converted to a probability distribution. In these cases, the distribution parameters are given as NA, but the fitted model itself is saved.

Usage

ddFit(x, ...)
# S3 method for data.frame
ddFit(
  x,
  distr = "standard",
  scVar = NULL,
  rCol = "r",
  expoCol = "exposure",
  ncarcCol = "ncarc",
  silent = FALSE,
  ...
)
# S3 method for rings
ddFit(
  x,
  distr = "standard",
  scVar = NULL,
  rCol = "r",
  expoCol = "exposure",
  ncarcCol = "ncarc",
  silent = FALSE,
  ...
)
# S3 method for list
ddFit(
  x,
  distr = "standard",
  scVar = NULL,
  rCol = "r",
  expoCol = "exposure",
  ncarcCol = "ncarc",
  silent = FALSE,
  ...
)
# S3 method for xyLayout
ddFit(
  x,
  distr = "standard",
  scVar = NULL,
  notSearched = NULL,
  rCol = "r",
  ncarcCol = "ncarc",
  unitCol = "turbine",
  silent = FALSE,
  ...
)
# S3 method for ringscc
ddFit(
  x,
  distr = "standard",
  scVar = NULL,
  rCol = "r",
  expoCol = "exposure",
  ncarcCol = "ncarc",
  silent = FALSE,
  ...
)

Value

A list of fitted glm models as dd objects in a

ddArray

object if a vector of distributions is fit, or a single dd

object if a single model is fit. The dd objects are lists that include the following elements:

glm: the fitted model
$distr: name of the distribution ("xep01", etc.)
$parms: vector of distribution parameter estimates (or NA if the model based on the MLE is not extensible)
$varbeta: the variance-covariance matrix of the glm parameter estimates. NOTE: This is identical to the covariance matrix from the glm, which can be extracted via summary(x)$cov.unscaled
$scVar: name of the (optional) search class variable (or NULL)
$ncarc: number of carcasses
$aicc: the AICc value of the fit
$n: number of rings
$k: number of parameters
$srad: search radius

When a dd object is printed, only a small subset of the elements are shown. To see a full list of the objects, use names(x). The elements can be extracted in the usual R way via $ or [[x]].

Arguments

x: a search plot layout object to fit carcass distribution models to. The layout may be a data frame with columns for ring radii, exposure (or searched area in each ring), search class variable (optional), and number of carcasses in each ring;
...: ignored
distr: names (vector of character strings) of glm distribution templates to fit. Default is distr = "standard" to fit the standard models listed in the description above. Setting distr = "all" will fit both the standard models and the supplementary models. Also, any subset of the models may be fit by using, for example, distr = c("xep01", "lognormal") to fit only the "xep01" and "lognormal" models, or distr = exclude(c("xep123", "constant")) to fit all standard models except "xep123" and "constant", or distr = exclude("lognormal", mod_all) to fit all the models except the lognormal.
scVar: Search class variable to include in the model (optional). scVar is ignored if x is not a shapeLayout or xyLayout object. If x is a shapeLayout object, scVar may be either NULL or the name of a single column with search class data. If x is an xyLayout object, scVar may be either NULL or a vector of names of search class variables to include in the models. If x is a data frame for a site "simple" layout turbines with varying detection probabilities among turbines, x must have a column named "type" to indicate a search plot type. The "type" for each turbine is a character string to differentiate turbine types that may have different detection probababilities (e.g., turbines 1 and 2 might be cleared plots with a relatively low carcass detection probability, while turbines 3 and 4 are searched on only the roads and pads with a relatively high carcass detection probability.
rCol: name of the distance column (which gives the outer radii of the rings). This will be correct by default for objects coming from prepRing and will rarely need to be explicitly specified.
expoCol: name of the column with the exposure, which is the area in the ring with outer radius rCol. This will be correct by default for objects coming from prepRing and will rarely need to be explicitly specified.
ncarcCol: name of the column with tallies of carcasses by ring. This will be correct by default for objects coming from prepRing and will rarely need to be explicitly specified.
silent: set silent = TRUE to suppress information printed to the console as the calculations proceed, which may be useful when running simulations.
notSearched: the name of the level (if any) in scVar that indicates an unsearched area
unitCol: name of the column with turbine IDs

Examples

Run this code

 data(layout_simple) 
 data(carcass_simple)
 sitedata <- initLayout(layout_simple) # initialize
 ringdata <- prepRing(sitedata) # format site layout data for modeling
 ringsWithCarcasses <- addCarcass(carcass_simple, data_ring = ringdata) # add carcasses to site
 distanceModels <- ddFit(ringsWithCarcasses) # fit distance models

Run the code above in your browser using DataLab