Create a set of mixed effects models

A MixedEffectsSet represents a group of mixed-effects models that all have the same functional structure. Fitting a large family of models (e.g., for many different species) using the same functional structure is a common pattern in allometric studies, and MixedEffectsSet facilitates the installation of these groups of models by allowing the user to specify the parameter estimates and descriptions in a dataframe or spreadsheet.

MixedEffectsSet(
  response,
  covariates,
  parameter_names,
  predict_fn,
  model_specifications,
  predict_ranef,
  fixed_only = FALSE,
  descriptors = list(),
  response_definition = NA_character_,
  covariate_definitions = list()
)

Arguments

response

A named list containing one element, with a name representing the response variable and a value representing the units of the response variable using the units::as_units function.

covariates

A named list containing the covariate specifications, with names representing the covariate name and the values representing the units of the coavariate using the units::as_units function

parameter_names

A character vector naming the columns in model_specifications that represent the parameters

predict_fn

A function that takes the covariate names as arguments and returns a prediction of the response variable. This function should be vectorized.

model_specifications

A dataframe such that each row of the dataframe provides model-level descriptors and parameter estimates for that model. Models must be uniquely identifiable using the descriptors. This is usually established using the load_parameter_frame() function.

predict_ranef

A function that predicts the random effects, takes any named covariates in covariates as arguments

fixed_only

A boolean value indicating if the model produces predictions using only fixed effects. This is useful when publications do not provide sufficient information to predict the random effects.

descriptors

An optional named list of descriptors that describe the context of the allometric model

response_definition

A string containing an optional custom response definition, which is used instead of the description given by the variable naming system.

covariate_definitions

An optional named list of custom covariate definitions that will supersede the definitions given by the variable naming system. The names of the list must match the covariate names given in covariates.

Value

An instance of MixedEffectsSet

Details

Because mixed-effects models already accommodate a grouping structure, MixedEffectsSet tends to be a much rarer occurrence than FixedEffectsSet and MixedEffectsModel.

Slots

parameters

A named list of parameters and their values

predict_fn_populated

The prediction function populated with the parameter values

specification

A tibble::tbl_df of the model specification, which are the parameters and the descriptors together

predict_ranef

The function that predicts the random effects

predict_ranef_populated

The function that predicts the random effects populated with the fixed effect parameter estimates

fixed_only

A boolean value indicating if the model produces predictions using only fixed effects

model_specifications

A tibble::tbl_df of model specifications, where each row reprents one model identified with descriptors and containing the parameter estimates.

Examples

mixed_effects_set <- MixedEffectsSet(
  response = list(
    vsia = units::as_units("ft^3")
  ),
  covariates = list(
    dsob = units::as_units("in")
  ),
  parameter_names = "a",
  predict_ranef = function(dsob, hst) {
    list(a_i = 1)
  },
  predict_fn = function(dsob) {
    (a + a_i) * dsob^2
  },
  model_specifications = tibble::tibble(a = c(1, 2))
)