mcmcBatscheletMixture: MCMC sampling for Batschelet-type distributions.
In keesmulder/flexcircmix: Fit Mixtures of Flexible Circular Distributions

Description Usage Arguments Value Examples

MCMC sampling for Batschelet-type distributions.

mcmcBatscheletMixture(x, Q = 1000, burnin = 0, thin = 1,
  n_comp = 4, bat_type = "inverse", init_pmat = matrix(NA, n_comp,
  4), fixed_pmat = matrix(NA, n_comp, 4), joint_kp_lam = FALSE,
  kp_bw = 1, lam_bw = 0.05, mu_logprior_fun = function(mu) -log(2 *
  pi), kp_logprior_fun = function(kp) 1,
  lam_logprior_fun = function(lam) -log(2), alph_prior_param = rep(1,
  n_comp), compute_variance = TRUE, compute_waic = FALSE,
  verbose = 0)

`x`	A numeric vector of angles, in radians
`Q`	Integer; The number of iterations to return after taking burn in and thinning into account.
`burnin`	Integer; The number of (non-thinned) iterations to discard. No burn in is performed by default.
`thin`	Integer; Number of iterations to sample for each saved iteration. Defaults to 1, which means no thinning.
`n_comp`	Integer; Fixed number of components to estimate.
`bat_type`	Either 'inverse' or 'power', the type of distribution to fit. The two distributions are similar, but the power Batschelet distribution is computationally much less demanding.
`init_pmat`	A numeric matrix with `n_comp` rows and four columns, corresponding to μ, κ, λ, α, in that order. Gives starting values for the parameters. If any element is `NA`, it will be given a default starting value. For mu, the default starting values are equally spaced on the circle. For κ, the default starting value is 5. For λ, the default starting value is 0, which corresponds to the von Mises distribution. For α, the default starting value is `1/n_comp`.
`fixed_pmat`	A numeric matrix with `n_comp` rows and four columns, corresponding to μ, κ, λ, α, in that order. Any element that is not `NA` in this matrix will be held constant at the given value and not sampled.
`joint_kp_lam`	Logical; If `TRUE`, the parameters `kp` and `lam` are drawn jointly. This can be beneficial if these are strongly correlated.
`kp_bw`	Numeric; A tuning parameter for kappa proposals. If `kp_bw == 1`, the chi-square distribution is used. Often, this distribution is too wide, so this parameter can be set to `0 < kp_bw < 1` to use a gamma proposal which has lower variance than the chi-square.
`lam_bw`	Numeric; the maximum distance from the current lambda at which uniform proposals are drawn.
`mu_logprior_fun`	Function; A function with a single argument, which returns the log of the prior probability of μ. Defaults to a uniform prior function.
`kp_logprior_fun`	Function; A function with a single argument, which returns the log of the prior probability of κ. Defaults to a uniform prior function. In contrast to the other parameters, for κ the constant (uniform) prior is improper.
`lam_logprior_fun`	Function; A function with a single argument, which returns the log of the prior probability of λ. Defaults to a uniform prior function.
`alph_prior_param`	Integer vector; The mixture weight parameter vector α is given its conjugate Dirichlet prior. The default is `rep(1, n_comp)`, which is the noninformative uniform prior over the `n_comp` simplex.
`compute_variance`	Logical; Whether to add circular variance to the returned mcmc sample.
`compute_waic`	Logical; Whether to compute the WAIC. Can be computationally demanding if `n * Q` is large.
`verbose`	Integer up to 4; Determines the amount of printed debug information.