Package 'BayesGWQS'

Bayesian Grouped WQS Regression

Description

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

Usage

bgwqs.fit(
  y,
  x,
  z,
  x.s,
  n.quantiles = 4,
  working.dir,
  n.chains = 1,
  n.iter = 10000,
  n.burnin = 5000,
  n.thin = 1,
  n.adapt = 500,
  DIC = FALSE
)

Arguments

y	A vector containing outcomes.
x	A matrix of component data.
z	A vector or matrix of controlling covariates.
x.s	A vector of the number of components in each index.
n.quantiles	The number of quantiles to apply to the component data.
working.dir	A file path to the directory.
n.chains	The number of Markov chains; must be a positive integer.
n.iter	The number of total iterations per chain, including burn in.
n.burnin	The number of iterations to discard at the beginning.
n.thin	The thinning rate; must be a positive integer.
n.adapt	The number of adaption iterations.
DIC	Logical; whether or not the user desires the function to return DIC.

Value

A list which includes BUGS output, sample chains post-burnin, and convergence test results.

Examples

## Not run: 
data("simdata")
group_list <- list(c("pcb_118", "pcb_138", "pcb_153", "pcb_180", "pcb_192"),
                   c("as", "cu", "pb", "sn"),
                   c("carbaryl", "propoxur", "methoxychlor", "diazinon", "chlorpyrifos"))
x.s <- make.x.s(simdata, 3, group_list)
X <- make.X(simdata, 3, group_list)
Y <- simdata$Y
work_dir <- tempdir()
results <- bgwqs.fit(y = Y, x = X, x.s = x.s, n.quantiles=4,
                     working.dir = work_dir,
                     n.chains = 1, n.iter = 10000, n.burnin = 5000, n.thin = 1, n.adapt = 500)


## End(Not run)

---

Forms matrix of components

Description

This function returns a matrix of component variables, X. The user can specify the desired chemicals and order by creating a list of string vectors, each vector containing the variable names of all desired elements of that group.

Usage

make.X(df, num.groups, groups)

Arguments

df	A dataframe containing named component variables
num.groups	An integer representing the number of component groups desired
groups	A list, each item in the list being a string vector of variable names for one component group

Value

A matrix of component variables

Examples

data("simdata")
group_list <- list(c("pcb_118", "pcb_138", "pcb_153", "pcb_180", "pcb_192"),
                   c("as", "cu", "pb", "sn"),
                   c("carbaryl", "propoxur", "methoxychlor", "diazinon", "chlorpyrifos"))
X <- make.X(simdata, 3, group_list)
X

---

Forms component group ID vector of X

Description

This function returns a vector which lets WQS.fit know the size and order of groups in X

Usage

make.x.s(df, num.groups, groups)

Arguments

df	A dataframe containing named component variables
num.groups	An integer representing the number of component groups desired
groups	A list, each item in the list being a string vector of variable names for one component group

Value

A vector of integers, each integer relating how many columns are in each group

Examples

data("simdata")
group_list <- list(c("pcb_118", "pcb_138", "pcb_153", "pcb_180", "pcb_192"),
                   c("as", "cu", "pb", "sn"),
                   c("carbaryl", "propoxur", "methoxychlor", "diazinon", "chlorpyrifos"))
x.s <- make.x.s(simdata, 3, group_list)
x.s

---

Simulated data of chemical concentrations and one binary outcome variable

Description

Data were simulated to have 0.7 in-group correlation and 0.3 between-group correlation. There are three groups, with the third being significantly correlated to the outcome variable.

Usage

simdata

Format

A data frame with 1000 rows and 15 variables:

pcb_118

a numeric vector; part of group 1

pcb_138

a numeric vector; part of group 1

pcb_153

a numeric vector; part of group 1

pcb_180

a numeric vector; part of group 1

pcb_192

a numeric vector; part of group 1

as

a numeric vector; part of group 2

cu

a numeric vector; part of group 2

pb

a numeric vector; part of group 2

sn

a numeric vector; part of group 2

carbaryl

a numeric vector; part of group 3

propoxur

a numeric vector; part of group 3

methoxychlor

a numeric vector; part of group 3

diazinon

a numeric vector; part of group 3

chlorpyrifos

a numeric vector; part of group 3

Y

a numeric vector; the outcome variable

---

Generates Plots of weights by group

Description

This function takes the object created by the bgwqs.fit function and a vector of group names and generates a random forest variable importance plot for each group. The weights in each group are listed in descending order.

Usage

weight.plot(fit.object, group.names, group.list, x.s)

Arguments

fit.object	The object that is returned by the bgwqs.fit function
group.names	A string vector containing the name of each group included in the BGWQS regression. Will be used for plot titles.
group.list	A list, each item in the list being a string vector of variable names for one component group.
x.s	A vector of the number of components in each index.

Value

A plot for each group of the BGWQS regression

---