contrast_weights: Build Contrast Weights

In mattansb/MSBMisc: Some functions I wrote that I find useful

Build Contrast Weights

Description

To be used ideally with emmeans::contrast. Each contrasts is tested (sum to 0?) and scaled so that all positive weights sum to 1 (and all negative weights to -1).

Usage

contrast_weights(..., .name = "custom", .adjust = NULL)

cw(..., .name = "custom", .adjust = NULL)

Arguments

...	Can be: Unnamed scalars. (Possibly named) vectors if equal length
.name	The label as it will appear in the results in emmeans.
.adjust	Gives the default adjustment method for multiplicity (used in emmeans).

Value

Depending on input, either a vector or a data frame of scaled weights.

Examples

data(mtcars)

mod <- lm(mpg ~ factor(cyl) * am, mtcars)


my_contrasts <- data.frame(
  "squares" = c(-1, 2, -1),
  "4 vs 6" = c(-30, 30, 0),
  check.names = FALSE
)

(my_contrasts2 <- cw(my_contrasts))
my_contrasts3 <- cw(my_contrasts, .adjust = "fdr")


library(emmeans)
(emms <- emmeans(mod, ~ cyl + am))

contrast(emms, method = my_contrasts, by = "am")
contrast(emms, method = my_contrasts2, by = "am") # estimate is affected!
contrast(emms, method = my_contrasts3, by = "am") # p value is affected

# Also in interaction contrasts
contrast(emms, interaction = list(cyl = my_contrasts2, am = "pairwise"))

mattansb/MSBMisc documentation built on March 22, 2023, 6:02 p.m.