fims_distributions::MultinomialLPMF< Type > Struct Template Reference

Implements the MultinomialLPMF distribution functor used by FIMS to evaluate the observation-level and total log-likelihood contributions under a multinomial error model for data, priors, and random effects. More...

#include <multinomial_lpmf.hpp>

Inheritance diagram for fims_distributions::MultinomialLPMF< Type >:

Public Member Functions
	MultinomialLPMF ()
	Constructor.
virtual	~MultinomialLPMF ()
	Destructor.
virtual const Type	evaluate ()
	Evaluates the multinomial log probability mass function.
Public Member Functions inherited from fims_distributions::DensityComponentBase< Type >
Type &	get_observed (size_t i)
	Retrieve one observed value based on input_type.
Type &	get_observed (size_t i, size_t j)
	Retrieve one observed matrix-like value based on input_type.
Type &	get_expected (size_t i)
	Retrieve one expected value based on input_type and use_mean.
size_t	get_n_x ()
	Get length of the active observed input vector.
size_t	get_n_expected ()
	Get length of the active expected input vector.
	DensityComponentBase ()
	Constructor, which initializes default prior pointer state and ID.
Public Member Functions inherited from fims_model_object::FIMSObject< Type >
uint32_t	GetId () const
	Getter that returns the unique id for parameters in the model.
bool	CheckDimensions (size_t actual, size_t expected)
	Check the dimensions of an object.

Public Attributes
fims::Vector< size_t >	dims
	Dimensions of the number of rows and columns of the multivariate dataset.
Public Attributes inherited from fims_distributions::DensityComponentBase< Type >
std::string	input_type
	Classification of the input pathway for this distribution object. Options used by accessor methods are, "prior", "random_effects", and "data".
std::shared_ptr< fims_data_object::DataObject< Type > >	data_observed_values
	Observed data.
fims::Vector< Type >	expected_values
	Expected value vector for prior-based pathways.
fims::Vector< Type > *	re = NULL
	Pointer to random effects vector.
fims::Vector< Type > *	re_expected_values = NULL
	Expected value vector for random-effects pathways.
fims::Vector< Type > *	data_expected_values = NULL
	Expected value vector for data pathways.
std::vector< fims::Vector< Type > * >	priors
	Vector of pointers where each entry points to a prior parameter.
fims::Vector< Type >	observed_values
	Input value of distribution function for priors or random effects.
fims::Vector< Type >	expected_mean
	The expected mean of the distribution; overrides expected values.
std::string	use_mean = fims::to_string("no")
	If "yes", expected_mean is used instead of expected_values. The default is "no" leading to the use of expected_values.
Type	lpdf
	Total log probability density contribution of the distribution.
int	observed_data_id_m = -999
	ID of observed data component.
fims::Vector< Type >	lpdf_vec
	Vector storing observation-level log-likelihood contributions.
bool	osa_flag = false
	Boolean; if true, one-step-ahead (OSA) residuals are calculated.
bool	simulate_flag = false
	Boolean; if true, data are simulated from the distribution.
std::vector< uint32_t >	key
	Unique ID for variable map that points to a fims::Vector.
Public Attributes inherited from fims_model_object::FIMSObject< Type >
uint32_t	id
std::vector< Type * >	parameters
std::vector< Type * >	random_effects_parameters
std::vector< Type * >	fixed_effects_parameters

Additional Inherited Members
Static Public Attributes inherited from fims_distributions::DensityComponentBase< Type >
static uint32_t	id_g = 0
	Global unique identifier for distribution modules.

Detailed Description

template<typename Type>
struct fims_distributions::MultinomialLPMF< Type >

Implements the MultinomialLPMF distribution functor used by FIMS to evaluate the observation-level and total log-likelihood contributions under a multinomial error model for data, priors, and random effects.

This implementation relies on TMB's R-style dmultinom() utility to compute row-wise multinomial log-probability mass contributions from observed counts (x_vector) and expected proportions (prob_vector). Specifically, when evaluating the multinomial likelihood, observations are passed to dmultinom(..., give_log = true).

For data input, if any element in a row is equal to na_value, the entire row is skipped and contributes zero to the objective. Contributions are stored in lpdf_vec. The summed total is returned by evaluate() and stored in lpdf.

Row observations could be counts of each age for a given time step, where additional time steps would be additional rows. Thus, columns are bins.

Member Function Documentation

evaluate()

template<typename Type >

virtual const Type fims_distributions::MultinomialLPMF< Type >::evaluate ( )

inlinevirtual

Evaluates the multinomial log probability mass function.

The following equation is the multinomial probability mass function, and thus, the log of it is evaluated:

\[ f(\underline{y}) = \frac{n!}{y_{1}!... y_{k}!}p^{y_{1}}_{1}...p^{y_{k}}_{k}, \]

where \(k\) is the number of categories, \(n\) is the sample size, \(\mu_{i}\) is the mean of \(y_{i}\) and is equal to \(np_{i}\), and \(\sigma^{2}_{i}\) is the variance of \(y_{i}\) and is equal to \(np_{i}(1-p_{i})\).

< total log probability mass contribution of the distribution

< index for lpdf_vec vector

Implements fims_distributions::DensityComponentBase< Type >.

---

The documentation for this struct was generated from the following file:

• inst/include/distributions/functors/multinomial_lpmf.hpp