covariates: Transforming Covariates and Filling in Missing Values

In nyiuab/BhGLM: Bayesian hierarchical GLMs and survival models, with applications to Genomics and Epidemiology

Transforming Covariates and Filling in Missing Values

Description

This function standardizes continuous variables, transforms categorical variables to indicator variables and centers the transformed variables.

Usage

covariates(x.con, x.cat, con.rescale = TRUE, cat.center = FALSE, fill.missing = TRUE) 

Arguments

x.con	a matrix or data frame consisting of continuous covariates.
x.cat	a matrix or data frame consisting of categorical covariates.
con.rescale	if con.rescale = TRUE, standardize continuous variables.
cat.center	if cat.center = TRUE, center categorical variables.
fill.missing	If fill.missing = TRUE, missing values of a variable are replaced by the mean of the observed data.

Value

This function returns a data frame containing the rescaled variables.

Note

For Bayesian hierarchical or penalization modeling, it is important to transform all variables to have a common scale before fitting data.

Author(s)

Nengjun Yi, [e-mail]

Examples

# fake data
age = rnorm(50, 30, 0.1)
sex = sample(x = c("male", "female"), size = 50, replace = T, prob = c(0.5, 0.5))
diet = sample(x = c(1, 5), size = 50, replace = T, prob = c(0.3, 0.7))
race = sample(c("Asian", "White", "Black"), size = 50, replace = T, prob = c(0.2, 0.4, 0.4))

x.con = age # continuous variable
x.cat = cbind(sex, diet, race) # categorical variables

x.con[1:2] = NA
x.cat[1,] = NA

x1 = covariates(x.con = x.con, x.cat = x.cat, con.rescale = F, cat.center = F, 
                fill.missing = T)
x1
x2 = covariates(x.con = x.con, x.cat = x.cat, con.rescale = T, cat.center = F, 
                fill.missing = T)
x2
x3 = covariates(x.con = x.con, x.cat = x.cat, con.rescale = T, cat.center = T, 
                fill.missing = T)
x3

nyiuab/BhGLM documentation built on June 12, 2024, 9:28 p.m.