Create and define a list of control options used to run the LDATS model, as implemented by LDA_TS.

LDA_TS_control(quiet = FALSE, measurer_LDA = AIC, selector_LDA = min,
  iseed = 2, memoise = TRUE, response = "gamma", lambda = 0,
  measurer_TS = AIC, selector_TS = min, ntemps = 6,
  penultimate_temp = 2^6, ultimate_temp = 1e+10, q = 0,
  nit = 10000, magnitude = 12, burnin = 0, thin_frac = 1,
  summary_prob = 0.95, seed = NULL, ...)

Arguments

quiet

logical indicator of whether the model should run quietly.

measurer_LDA, selector_LDA

Function names for use in evaluation of the LDA models. measurer_LDA is used to create a value for each model and selector_LDA operates on the values to choose the model.

iseed

integer initial seed for the LDA model set.

memoise

logical indicator of whether the multinomial functions should be memoised (via memoise). Memoisation happens to both multinom_TS and multinom_TS_chunk.

response

character element indicating the response variable used in the time series. Should be set to "gamma" for LDATS.

lambda

numeric "weight" decay term used to set the prior on the regressors within each chunk-level model. Defaults to 0, corresponding to a fully vague prior.

measurer_TS, selector_TS

Function names for use in evaluation of the TS models. measurer_TS is used to create a value for each model and selector_TS operates on the values to choose the model.

ntemps

integer number of temperatures (chains) to use in the ptMCMC algorithm.

penultimate_temp

Penultimate temperature in the ptMCMC sequence.

ultimate_temp

Ultimate temperature in the ptMCMC sequence.

q

Exponent controlling the ptMCMC temperature sequence from the focal chain (reference with temperature = 1) to the penultimate chain. 0 (default) implies a geometric sequence. 1 implies squaring before exponentiating.

nit

integer number of iterations (steps) used in the ptMCMC algorithm.

magnitude

Average magnitude (defining a geometric distribution) for the proposed step size in the ptMCMC algorithm.

burnin

integer number of iterations to remove from the beginning of the ptMCMC algorithm.

thin_frac

Fraction of iterations to retain, from the ptMCMC. Must be \((0, 1]\), and the default value of 1 represents no thinning.

summary_prob

Probability used for summarizing the posterior distributions (via the highest posterior density interval, see HPDinterval) of the TS model.

seed

Input to set.seed in the time series model for replication purposes.

...

Additional arguments to be passed to LDA as a control input.

Value

list of control lists, with named elements LDAcontrol, TScontrol, and quiet.

Examples

  LDA_TS_control()