Consider the following code where we fit a regression model, alter the data and create the corresponding design matrix for the altered data based on the fitted regression model:
mtcars_alt <- mtcars |>
dplyr::mutate(vs = factor(vs),
am = factor(am))
mod_lm <- lm(mpg ~ vs*am, data = mtcars_alt)
mtcars_alt |>
dplyr::mutate(am = "0") |>
model.matrix(mod_lm, data = _)
mod_glm <- glm(mpg ~ vs*am, data = mtcars_alt)
mtcars_alt |>
dplyr::mutate(am = "0") |>
model.matrix(mod_glm, data = _)
When the regression model is based on lm() then it works fine, however, it throws an error when it's based on glm()
Error in `contrasts<-`(`*tmp*`, value = contr.funs[1 + isOF[nn]]) :
contrasts can be applied only to factors with 2 or more levels
Any idea how to circumvent this error?
If you call model.matrix() on the GLM's formula rather than the model object itself and create the constant factor am as a factor with two levels (even though only one is observed, as suggested by @Stefan Hansen in the comments), you should get the desired result.
mtcars_alt <- mtcars |>
dplyr::mutate(vs = factor(vs),
am = factor(am))
mod_glm <- glm(mpg ~ vs*am, data = mtcars_alt)
mtcars_alt |>
dplyr::mutate(am = factor("0", levels=c("0", "1"))) |>
model.matrix(formula(mod_glm), data = _)
#> (Intercept) vs1 am1 vs1:am1
#> Mazda RX4 1 0 0 0
#> Mazda RX4 Wag 1 0 0 0
#> Datsun 710 1 1 0 0
#> Hornet 4 Drive 1 1 0 0
#> Hornet Sportabout 1 0 0 0
#> Valiant 1 1 0 0
#> Duster 360 1 0 0 0
#> Merc 240D 1 1 0 0
#> Merc 230 1 1 0 0
#> Merc 280 1 1 0 0
#> Merc 280C 1 1 0 0
#> Merc 450SE 1 0 0 0
#> Merc 450SL 1 0 0 0
#> Merc 450SLC 1 0 0 0
#> Cadillac Fleetwood 1 0 0 0
#> Lincoln Continental 1 0 0 0
#> Chrysler Imperial 1 0 0 0
#> Fiat 128 1 1 0 0
#> Honda Civic 1 1 0 0
#> Toyota Corolla 1 1 0 0
#> Toyota Corona 1 1 0 0
#> Dodge Challenger 1 0 0 0
#> AMC Javelin 1 0 0 0
#> Camaro Z28 1 0 0 0
#> Pontiac Firebird 1 0 0 0
#> Fiat X1-9 1 1 0 0
#> Porsche 914-2 1 0 0 0
#> Lotus Europa 1 1 0 0
#> Ford Pantera L 1 0 0 0
#> Ferrari Dino 1 0 0 0
#> Maserati Bora 1 0 0 0
#> Volvo 142E 1 1 0 0
#> attr(,"assign")
#> [1] 0 1 2 3
#> attr(,"contrasts")
#> attr(,"contrasts")$vs
#> [1] "contr.treatment"
#>
#> attr(,"contrasts")$am
#> [1] "contr.treatment"
Created on 2024-11-14 with reprex v2.1.0