zoid (version 1.3.1)

fit_zoid: Fit a trinomial mixture model with Stan

Description

Fit a trinomial mixture model that optionally includes covariates to estimate effects of factor or continuous variables on proportions.

Usage

fit_zoid(
  formula = NULL,
  design_matrix,
  data_matrix,
  chains = 3,
  iter = 2000,
  warmup = floor(iter/2),
  overdispersion = FALSE,
  overdispersion_sd = 5,
  posterior_predict = FALSE,
  moment_match = FALSE,
  prior_sd = NA,
  ...
)

Arguments

formula

The model formula for the design matrix. Does not need to have a response specified. If =NULL, then the design matrix is ignored and all rows are treated as replicates

design_matrix

A data frame, dimensioned as number of observations, and covariates in columns

data_matrix

A matrix, with observations on rows and number of groups across columns

chains

Number of mcmc chains, defaults to 3

iter

Number of mcmc iterations, defaults to 2000

warmup

Number iterations for mcmc warmup, defaults to 1/2 of the iterations

overdispersion

Whether or not to include overdispersion parameter, defaults to FALSE

overdispersion_sd

Prior standard deviation on 1/overdispersion parameter, Defaults to inv-Cauchy(0,5)

posterior_predict

Whether or not to return draws from posterior predictive distribution (requires more memory)

moment_match

Whether to do moment matching via loo::loo_moment_match(). This increases memory by adding all temporary parmaeters to be saved and returned

prior_sd

Parameter to be passed in to use as standard deviation of the normal distribution in transformed space. If covariates are included this defaults to 1, but for models with single replicate, defaults to 1/n_bins.

...

Any other arguments to pass to rstan::sampling().

Examples

# \donttest{
y <- matrix(c(3.77, 6.63, 2.60, 0.9, 1.44, 0.66, 2.10, 3.57, 1.33),
  nrow = 3, byrow = TRUE
)
# fit a model with no covariates
fit <- fit_zoid(data_matrix = y, chains = 1, iter = 100)

# fit a model with 1 factor
design <- data.frame("fac" = c("spring", "spring", "fall"))
fit <- fit_zoid(formula = ~fac, design_matrix = design, data_matrix = y, chains = 1, iter = 100)
# }
# try a model with random effects
set.seed(123)
y <- matrix(runif(99,1,4), ncol=3)
design <- data.frame("fac" = sample(letters[1:5], size=nrow(y), replace=TRUE))
design$fac <- as.factor(design$fac)
fit <- fit_zoid(formula = ~(1|fac), design_matrix = design, data_matrix = y, chains = 1, iter = 100)