Create forest plot of confidence intervals

"Forest plot"-style plotting of confidence intervals from a regression model. Basic input is a matrix with columns of estimate/lower/upper, along with an optional 4th column for the p-value. Also works with a variety of models (lm/glm/coxph/etc).

Usage

ci_plot(obj, ...)

# S3 method for class 'matrix'
ci_plot(
  obj,
  sort = TRUE,
  diff = (ncol(obj) == 4),
  null = 0,
  trans,
  p_label = FALSE,
  ...
)

# S3 method for class 'lm'
ci_plot(obj, intercept = FALSE, exclude = NULL, plot = TRUE, tau, ...)

# S3 method for class 'glm'
ci_plot(obj, ...)

# S3 method for class 'mer'
ci_plot(
  obj,
  intercept = FALSE,
  exclude = NULL,
  plot = TRUE,
  tau,
  nsim = 500,
  ...
)

# S3 method for class 'coxph'
ci_plot(obj, exclude = NULL, plot = TRUE, tau, ...)

# S3 method for class 'data.frame'
ci_plot(obj, ...)

Arguments

obj

The object to be plotted; can be a matrix of raw values or a model object

...

Not used

sort

Sort parameters by estimate? (default: true)

diff

Include tests of difference / p-values?

null

Draw a line representing no effect at this value (default: 0)

trans

Transformation to be applied (e.g., trans=exp for a log link)

p_label

Label p-values (p=0.02 instead of just 0.02)? (default: FALSE)

intercept

Include a CI for the intercept? (default: FALSE)

exclude

Variables to exclude (character vector)

plot

If FALSE, just returns the matrix of estimates/CIs/p-values to be plotted but doesn't plot anything

tau

A named vector of effect sizes; CIs will be shown for tau*beta. Any coefficients not included are given tau = 1.

nsim

Number of simulations; see confint.merMod

Details

TO-DO:

• interactions

See also

ci_int()

Examples

# Supplying a matrix
B <- cbind(1:9, 0:8, 2:10)
rownames(B) <- LETTERS[1:9]
ci_plot(B)


# Supplying a fitted model object
fit <- lm(Ozone ~ Solar.R + Wind + Temp, airquality)
ci_plot(fit)

ci_plot(fit, tau=c(Solar.R = 100, Temp = 10, Wind = 3))

ci_plot(fit, tau=c(Solar.R = 100))