Compute the min-log-likelihood

Use data to compute the min-log-likelihood of choosing a given observed cell given a set of parameters. Should allow the user to estimate the parameters of the model.

Usage

mll(
  data,
  parameters,
  parameter_names = colnames(params_from_csv[["params_archetypes"]])[-c(1, 2)],
  transform = TRUE,
  bounds = params_from_csv[["params_bounds"]],
  cpp = TRUE,
  summed = FALSE,
  ...
)

Arguments

data

Data.frame containing at least "id", "time", "x", "y", "goal_x", "goal_y", and "goal_id". If it does not have the utility variables yet, these will add them to the data.frame.

parameters

Numeric vector or matrix containing the parameters to be used. Should be specified in the same order as specified in "parameter_names". If a matrix, each row should contain parameters to be estimated for each instance of "id" separately.

parameter_names

Character vector containing the parameters that you want to estimate. Defaults to all parameters defined in params_from_csv. Whenever not all parameters are used, the excluded parameters are assumed to have a value of 0.

transform

Logical denoting whether to transform the provided parameters from the real axis to the bounded scales imposed on the parameters within predped. Defaults to TRUE.

bounds

Matrix containing the lower and upper bounds of the parameters in its first and second column respectively. Additionally, rownames should denote for which parameter a certain pair represents the bounds. Only used when transform = TRUE. Defaults to the default bounds of predped.

cpp

Logical denoting whether to use the mll_rcpp function to compute the min-log-likelihood. Defaults to TRUE.

summed

Logical denoting whether to sum the min-log-likelihood to one value per person. If TRUE, you get the resulting summed min-log-likelihood for each individual with a correction to avoid -Infs. If FALSE, the function will instead return a list of vectors containing the raw likelihoods (not min-log-likelihoods!), allowing users to specify their own corrections (if needed). Defaults to FALSE.

...

Additional arguments passed on to add_motion_variables. In a typical estimation situation, these motion variables should already be in data.

@return Either named vector containing the summed min-log-likelihood (summed = TRUE) or named list with vectors of raw likelihoods (summed = FALSE) per person in the dataset.