Get model means and slopes (trends) — model

Retrieve estimated marginal means and slopes (trends) as computed by emmeans::emmeans and emmeans::emtrends, and tests for all pairwise differences, as computed by emmeans::pairs.emmGrid.

Usage

model_means(
  model,
  formula,
  cld,
  backtransform = TRUE,
  type = if (isTRUE(backtransform)) "response" else "linear",
  ...
)

pairwise_model_means(
  model,
  formula,
  backtransform = TRUE,
  type = if (isTRUE(backtransform)) "response" else "linear",
  ...
)

model_slopes(model, formula, ..., cld = TRUE)

pairwise_model_slopes(model, formula, ...)

Arguments

model: a linear model or generalized linear model.
formula: the desired means or slopes; see Details.
cld: a logical variable specifying if a compact letter display should be used for pairwise comparisons between groups. Defaults to TRUE unless the desired means are for a numerical variable or there is only one mean reported.
backtransform: if a linear model and response variable is of form log(y), or the model is a logistic model , backtransform the resulting estimate and confidence interval bounds, so that they report either geometric means on the original scale and ratios (for log-transformed responses) or proportions and odds ratios (for logistic models).
type: specify method of backtransformation to emmeans. If used, backtransform is ignored.
...: additional parameters, passed to emmeans or emtrends. See Details.

Details

For means, specify the desired combinations of explanatory variables using formula notation. For example, to get the means at all levels of a variable x, use ~ x. To get the means at all combinations of x1 and x2, use ~ x1 + x2.

For slopes, specify the desired slope on the left side of the formula. For example, to get the slopes for a variable x at each level of a variable g, use x ~ g. To get the overall slope, use x ~ 1.

Additionally, one can specify groupings by using a | within a formula. For example, to get the means of all combinations of x1 and x2 grouped by each value of x2, use ~ x1 | x2. This is especially useful for getting pairwise tests of means and trends within subgroups.

By default, estimated marginal means and trends are estimated by computing the average mean or trend across all values of any other categorical explanatory variables equally, and at the mean value of any other numeric explanatory variables.

One of most useful additional parameters that can be passed to emmeans is at, which allows one to specify specific values of continuous variables to compute means and/or trends at. For example, if one had a continuous variable x in the model, one could compute the means at x=10 and x=20 using at = list(x = c(10, 20).