as.data.frame.dynamitefit: Extract Samples From a `dynamitefit` Object as a Data Frame

Description

Provides a data.frame representation of the posterior samples of the model parameters.

Usage

# S3 method for dynamitefit
as.data.frame(
  x,
  row.names = NULL,
  optional = FALSE,
  types = NULL,
  parameters = NULL,
  responses = NULL,
  times = NULL,
  groups = NULL,
  summary = FALSE,
  probs = c(0.05, 0.95),
  include_fixed = TRUE,
  ...
)

Value

A tibble containing either samples or summary statistics of the model parameters in a long format. For a wide format, see as_draws.dynamitefit().

Arguments

x: [dynamitefit]
The model fit object.
row.names: Ignored.
optional: Ignored.
types: [character()]
Type(s) of the parameters for which the samples should be extracted. See details of possible values. Default is all values listed in details except spline coefficients omega. This argument is mutually exclusive with parameters.
parameters: [character()]
Parameter(s) for which the samples should be extracted. Possible options can be found with function get_parameter_names(). Default is all parameters of specific type for all responses. This argument is mutually exclusive with types.
responses: [character()]
Response(s) for which the samples should be extracted. Possible options are elements of unique(x$priors$response), and the default is this entire vector. Ignored if the argument parameters is supplied. omega_alpha, and omega_psi. See also get_parameter_types().
times: [double()]
Time point(s) to keep. If NULL (the default), all time points are kept.
groups: [character()]
Group name(s) to keep. If NULL (the default), all groups are kept.
summary: [logical(1)]
If TRUE, returns posterior mean, standard deviation, and posterior quantiles (as defined by the probs argument) for all parameters. If FALSE (default), returns the posterior samples instead.
probs: [numeric()]
Quantiles of interest. Default is c(0.05, 0.95).
include_fixed: [logical(1)]
If TRUE (default), time-varying parameters for 1:fixed time points are included in the output as NA values. If FALSE, fixed time points are omitted completely from the output.
...: Ignored.

Details

The arguments responses and types can be used to extract only a subset of the model parameters (i.e., only certain types of parameters related to a certain response variable).

Potential values for the types argument are:

alpha
Intercept terms (time-invariant or time-varying).
beta
Time-invariant regression coefficients.
cutpoint
Cutpoints for ordinal regression.
delta
Time-varying regression coefficients.
nu
Group-level random effects.
lambda
Factor loadings.
kappa
Contribution of the latent factor loadings in the total variation.
psi
Latent factors.
tau
Standard deviations of the spline coefficients of delta.
tau_alpha
Standard deviations of the spline coefficients of time-varying alpha.
sigma_nu
Standard deviations of the random effects nu.
corr_nu
Pairwise within-group correlations of random effects nu. Samples of the full correlation matrix can be extracted manually as rstan::extract(fit$stanfit, pars = "corr_matrix_nu") if necessary.
sigma_lambda
Standard deviations of the latent factor loadings lambda.
corr_psi
Pairwise correlations of the noise terms of the latent factors. Samples of the full correlation matrix can be extracted manually as rstan::extract(fit$stanfit, pars = "corr_matrix_psi") if necessary.
sigma
Standard deviations of Gaussian responses.
corr
Pairwise correlations of multivariate Gaussian responses.
phi
Describes various distributional parameters, such as:
- Dispersion parameter of the Negative Binomial distribution.
- Shape parameter of the Gamma distribution.
- Precision parameter of the Beta distribution.
- Degrees of freedom of the Student t-distribution.
omega
Spline coefficients of the regression coefficients delta.
omega_alpha
Spline coefficients of time-varying alpha.
omega_psi
Spline coefficients of the latent factors psi. Note that in case of nonzero_lambda = FALSE, mean of these are used to flip the sign of psi to avoid multimodality due to sign-switching, but omega_psi variables are not modified.
zeta
Total variation of latent factors, i.e., the sum of sigma_lambda and tau_psi.

Examples

Run this code

data.table::setDTthreads(1) # For CRAN
as.data.frame(
  gaussian_example_fit,
  responses = "y",
  types = "beta"
)

# Basic summaries can be obtained automatically with summary = TRUE
as.data.frame(
  gaussian_example_fit,
  responses = "y",
  types = "beta",
  summary = TRUE
)

# Time-varying coefficients "delta"
as.data.frame(
  gaussian_example_fit,
  responses = "y",
  types = "delta",
  summary = TRUE
)

# Obtain summaries for a specific parameters
as.data.frame(
  gaussian_example_fit,
  parameters = c("tau_y_x", "sigma_y"),
  summary = TRUE
)