calc_nb_counts: Calculate predicted value and standard error of a negative...

In jenniferthompson/JTHelpers: Personal Helper Functions, Some Related to rms Package

Description

Calculate predicted value and standard error of a negative binomial regression model for each row in a design matrix.

Usage

calc_nb_counts(nbObj, designMatrix, predVar = NULL, intVar = NULL)

Arguments

nbObj	Model fit of class glm.nb, or mice::mira object fit using glm.nb.
designMatrix	Matrix of covariate values. Number of columns = number of coefficients in nbObj.
predVar	Character string; name of main predictor variable. Defaults to NULL, in which case predictor variable values will not be included in resulting data.frame.
intVar	Character string (optional); name of interacting variable. Defaults to NULL. If included, nothing changes except this column in designMatrix will also be included in the returned data.frame.

Value

data.frame containing columns for main predictor variable; interacting variable (if given); linear predictor and its SE; adjusted count and its lower and upper confidence limits.

See Also

glm.nb; is.mice, pool.

Examples

## Create data frame
df <- data.frame(y = round(rexp(n = 100, rate = 0.5)),
                 v1 = sample(1:5, size = 100, replace = TRUE),
                 v2 = rnorm(n = 100))

## Fit negative binomial model
mymod <- MASS::glm.nb(y ~ v1 * v2, data = df)

## Create design matrix
mydmat <- matrix(c(rep(1, 5), 1:5, rep(median(df$v2), 5)), ncol = 3)
mydmat <- cbind(mydmat, mydmat[,2] * mydmat[,3])
colnames(mydmat) <- c('(Intercept)', 'v1', 'v2', 'v1:v2')

calc_nb_counts(mymod, designMatrix = mydmat, predVar = 'v1', intVar = 'v2')

jenniferthompson/JTHelpers documentation built on May 19, 2019, 4:04 a.m.