bwqs_r: Fitting Bayesian Weighted Quantile Sum regression models
In ElenaColicino/bwqs: Bayesian Weighted Quantile Sum Regression

View source: R/bwqs_main-revised.R

bwqs_r

R Documentation

Fitting Bayesian Weighted Quantile Sum regression models

Description

Fits Random Bayesian Weighted Quantile Sum (BWQS) regressions for continuous outcomes. This model provides estimation for the mixture composition and overall effect of the mixture across different groups on the outcomes using bayesian framework.

Usage

bwqs_r(
  formula,
  mix_name,
  cluster_name,
  data,
  q,
  Dalp = NULL,
  chains = 1,
  iter = 1000,
  thin = 3,
  seed = 2019,
  start_value = NULL,
  c_int = c(0.025, 0.975),
  family = "gaussian"
)

Arguments

`formula`	Object of class `formula` specifying the relationship between the outcome and the covariates of the model not involved in the mixture variable. If the model has no covariates specify `y ~ NULL`.
`mix_name`	A character vector listing the variables contributing to a mixture effect.
`cluster_name`	A character string that specifiy which is the column of the dataset which contains the group number. Note that the `cluster_name` should be numeric, strings and factors are not allowed
`data`	The `data.frame` containing the variables (covariates and elements of the mixture) to be included in the model.
`q`	An `integer` to specify how mixture variables will be ranked, e.g. in quartiles (`q = 4`), deciles (`q = 10`), or percentiles (`q = 100`). If `q = NULL` then the values of the mixture variables are taken (these must be standardized or the domain must be the same).
`Dalp`	A `vector` containing the parameters of the Dirichlet distribution of the weights, the number of the elements of the vector has to be equal to the number of chemicals. If `Dalp = NULL` the domain is explored uniformly.
`chains`	An `integer` to specify the number of chain in Hamiltonian Monte Carlo algorithm. Default value `chains = 1`.
`iter`	An `integer` to specify the lenght of chain in Hamiltonian Monte Carlo algorithm. Default value `iter = 10000`.
`thin`	An `integer` to specify the thinning parameter in Hamiltonian Monte Carlo algorithm.
`seed`	An `integer` value to fix the seed. If `seed = NULL` the seed are randomly choosen.
`start_value`	A `vector` containing the initial value of the prior distribution, if it is equal to NULL random values are chosen.
`c_int`	A `vector` of two elements to specify the credible intervals for parameters, for 95% credible interval `c_int = c(0.025,0.975)` (default).
`family`	A `string` to specify the type of outcome. With the current implementation the possible values are only continuous - "gaussian" (default).

Details

The function bwqs uses the package rstan which allows the connection with STAN, a specific software, written in C++ for bayesian inference, for further information see https://mc-stan.org/.

Value

bwqs returns a list with two argument:

`fit`	An `S4` object with all details of the Hamiltonian Monte Carlo, all the extractions from the posterior distribution and all values of the parameters
`summary_fit`	Table with the statistics of the parameters: mean, standard error of the mean, standard deviation, lower and upper values for the credible interval (with credible level specified by `c_int`), n_eff and Rhat. For further details see https://cran.r-project.org/web/packages/rstan/rstan.pdf