circ_mle {CircMLE}

circ_mle

Description

Run all 10 maximum likelihood models of circular orientation

Usage

circ_mle(
  data,
  criterion = "AIC",
  nchains = 5,
  BadStart = 10^9,
  niter = 5000,
  method = "BFGS",
  lambda.min = 0.25,
  q.diff,
  exclude = NULL
)

Arguments

data

A vector of class 'circular'

criterion

chose from either "AIC", "AICc", or "BIC" for the model comparison information criterion. (default = "AIC")

nchains

A positive integer indicating the number of chains to run. Only the chain with the lowest log likelihood is returned (default = 5)

BadStart

An integer to replace the log likelihood when starting parameters of the optimizer fall outside the preset bounds. This is usually set to a large number. Can also be set to Inf, but will result in an error if a method other than "Nelder-Mead" is chosen. (default = 10^9)

niter

The maximum number of iterations for the optimizing algorithm. Equivalent to the 'maxit' control parameter of the optim() function. See ?optim for more details. (default = 5000)

method

A character string indicating the optimizing algorithm to use. Either "BFGS" or "Nelder-Mead" are recommended. See ?optim for more details. (default = "BFGS")

lambda.min

The minimum proportional size of the first distribution. Must be between 0 and 1. (default = 0.25)

q.diff

The minimum difference (in radians) in preferred direction for bimodal models. Must be set between 0 and pi. (default = pi/4)

exclude

A character vector of the models to be excluded from the calculations. (The default is to include all 10 models). For example, exclude = c("M1", "M3A", "M5B").

Value

A list with 4 elements:

$results: A data frame consisting of a row for each model (rownames) with the columns: 1 = number of free parameters, 2 = mu1, 3 = kappa1, 4 = lamda, 5 = mu2, 6 = kappa2, 7 = negative log likelihood, 8 = Counts.function, 9 = Counts.Gradient, 10 = convergence, 11 = message, 12 = AIC, 13 = AICc, 14 = BIC, 15 = delta AIC, 16 = delta AICc, 17 = delta BIC, 18 = relative likelihoods of criterion chosen, 19 = model weights (probabilities) for criterion chosen, 20 = evidence ratios for the best model selected by the criterion. See ?optim for details on columns 8-11.

$bestmodel: The best model according to the criterion chosen

$rt: A two-element vector giving the test statistic and p-value for the Rayleigh Test

$hessians: A list with each element containing the hessian matrix for each model. Used for calculating connfidence intervals of parameters.

Examples

testdata = circular::rvonmises(100, mu = circular::circular(pi), kappa = 3)
circ_mle(testdata)