rcsOR {interactionRCS}

Restricted cubic spline interaction OR for more than 3 knots

Description

Generate OR values in a logistic model for a 1 unit increase in a variable at specified points of another interacting variable splined with rcs(df >= 3)

Usage

rcsOR(
  var2values,
  model,
  data = NULL,
  var1,
  var2,
  ci = TRUE,
  conf = 0.95,
  ci.method = "delta",
  ci.boot.method = "perc",
  R = 100,
  parallel = "multicore",
  ...
)

Arguments

var2values

numeric vector of var2 points to estimate

model

model of class lrm, Glm or glm family binomial. If data is NULL, the function expects to find the data in model$x.

data

data used in the model. If absent, we will attempt to recover the data from the model. Only used for bootstrap and glm class models

var1

variable that increases by 1 unit from 0

var2

variable to spline. var2values belong to var2

ci

calculate 95% CI?

conf

confidence level. Default 0.95

ci.method

confidence interval method. "delta" performs delta method. "bootstrap" performs bootstrapped CI (slower)

ci.boot.method

one of the available bootstrap CI methods from boot.ci. Default percentile

R

number of bootstrap samples if ci.method = "bootstrap". Default 100

parallel

can take values "no", "multicore", "snow" if ci.method = "bootstrap". Default multicore

...

other parameters for boot

Value

if ci = FALSE, a dataframe with initial values and OR

Examples

library(rms)
library(mlbench)
data(PimaIndiansDiabetes)
# Set age on a 5-year scale
PimaIndiansDiabetes$age <- PimaIndiansDiabetes$age/5
# Recode diabetes as 0/1
PimaIndiansDiabetes$diabetes <- ifelse(PimaIndiansDiabetes$diabetes=="pos" , 1 , 0)
myformula <- diabetes ~ mass + age * rcs( glucose , 4 )
model <- glm(myformula , data = PimaIndiansDiabetes , family = "binomial")
rcsOR( var2values = 20:80
       , model = model , data = PimaIndiansDiabetes , var1 ="age", var2="glucose"
       , ci=TRUE , conf = 0.95 , ci.method = "delta")