lincom: Create estimates, CIs and P-values for a linear combination...

In bmckuw/UWbe536: UW BIOST/EPI 536 Utility Functions

Description

Create estimates, CIs and P-values for a linear combination of two coefficients

Usage

lincom(model, lc, robust = FALSE, expo = TRUE, digits = 3,
  type = "HC1", level = 0.95, df = Inf)

Arguments

model	The fit of a generalized linear model (lm or glm)
lc	Character vector for the linear combination(s) set equal to null value(s) (usually zero) to test. Written in terms of variable names.
robust	Logical expression indicating whether robust standard errors are to be used
expo	Logical expression indicating whether exponentiated linear combination of coefficients is requiested
digits	Number of digits to round to in output
type	Character string giving type of robust standard error from sandwich package
level	Numeric confidence coefficient for CIs (as proportion)
df	Degrees of freedom for reference t distribution (Inf means Normal distribution)

Value

A table of linear combination values or exponentiated linear combination values with CI(s)s (robust or not) and P-value(s).

Examples

x1 <- rbinom(100, 2, .5)
x2 <- rbinom(100, 2, .5)
expit <- function(x) exp(x)/(1 + exp(x))
y <- rbinom(100, 1, p = expit(x1 + x2 + .1))
model <- glm(y~x1*x2, family = "binomial")
variable.names(model)
lincom(model, lc = "x1 + x1:x2 == 0")

bmckuw/UWbe536 documentation built on May 31, 2019, 1:54 p.m.