gwqs.fit {groupWQS}

Grouped WQS Regression

Description

This function fits a grouped weighted quantile sum (GWQS) regression model.

Usage

gwqs.fit(
  y,
  y.train = NULL,
  x,
  x.train = NULL,
  z = NULL,
  z.train = NULL,
  x.s,
  B = 100,
  n.quantiles = 4,
  pars = NULL,
  func,
  ineqLB = NULL,
  ineqUB = NULL,
  tol = 1e-06,
  delta = 1e-06
)

Arguments

y

A vector containing outcomes for validation.

y.train

A vector containing outcomes for training. If left as NULL the validation data will be used for training as well.

x

A matrix of component data for validation.

x.train

A matrix of component data for training. If left as NULL the validation data will be used for training as well.

z

A vector or matrix of covariates for validation.

z.train

A vector or matrix of covariates for training. If left as NULL the validation data will be used for training as well.

x.s

A vector of the number of components in each index.

B

The number of bootstrap samples, must be 1 or more.

n.quantiles

The number of quantiles to apply to data.

pars

A vector of initial values, listed in order: beta naught intercept and group index beta coefficients, individual chemical weight coefficients, and covariate coefficients.

func

The objective function to be used (must match outcome data type); currently only fun args "continuous" or "binary" are supported.

ineqLB

Vector of lower bounds for betas and weights, set to -2 by default.

ineqUB

Vector of upper bounds for betas and weights, set to 2 be default.

tol

Tolerance level for bootstrap convergence.

delta

Step size for bootstrap procedure.

Value

A list of 3 containing the GWQS estimate based on calculated weights, the GWQS model fit to validation data, and weight estimates

Examples

data("WQSdata")
group_list <- list(c("X1", "X2", "X3"), c("X4", "X7"), c("X5", "X6", "X9", "X8"))
x.s <- make.x.s(WQSdata, 3, group_list)
X <- make.X(WQSdata, 3, group_list)
Y <- WQSdata$y
results <- gwqs.fit(y = Y, x = X, x.s = x.s, B=1, func = "continuous")