Native estimated marginal means (least-squares means) for lm
and glm fits, following the conventions of
emmeans::emmeans() (Lenth): a reference grid crossing every
level of each categorical predictor with each numeric covariate held
at its mean (or at the values in at); the marginal mean for a
cell of specs is the equal-weight average of the grid
predictions over the predictors not named in specs, a linear
combination \(L b\) with standard error \(\sqrt{L V L'}\),
\(V = \) vcov(model). Degrees of freedom are the residual df
for lm and gaussian / Gamma glm fits and Inf
(asymptotic z) otherwise. Results are on the link scale unless
type = "response", which back-transforms the estimate and the
interval and scales the SE by the inverse-link derivative (delta
method), naming the estimate prob (binomial), rate
(poisson) or response. Cross-validated against emmeans
in the tests; emmeans is not needed at run time.
Arguments
- model
A fitted
lmorglmobject.- specs
Variables whose marginal means are wanted: a one-sided formula (
~ g,~ g * h,~ g | h) or a character vector of predictor names.- ...
Optional native settings:
type("link", default, or"response"),level(confidence level, default 0.95),at(named list of covariate values for the grid) andby(conditioning variables). Any other argument is an error.
Value
A data frame shaped like summary(emmeans::emmeans(...)):
the spec (and by) variables, the estimate (emmean /
prob / rate / response), SE, df
and lower.CL/upper.CL (asymp.LCL/
asymp.UCL when df = Inf). Attributes carry the
estimate name, scale and the linear-function matrix
("linfct"), so a contrast matrix C gives contrasts as the matrix product of C and
attr(x, "linfct") times the coefficients.
The emmGrid S4 object is no longer returned.
Examples
set.seed(1)
df <- data.frame(y = rnorm(60), x = rnorm(60), g = factor(rep(c("a", "b"), 30)))
fit <- stats::lm(y ~ x + g, data = df)
morie_effects_emmeans(fit, specs = "g")
#> g emmean SE df lower.CL upper.CL
#> 1 a 0.2411587 0.1567827 57 -0.07279328 0.5551108
#> 2 b -0.0259260 0.1567827 57 -0.33987803 0.2880260