I am trying to compute predictions using the R::marginaleffects package. I would like to do this several times for specific predictor variables of interest, within a map / apply like function, but am struggling to get the syntax right. My attempt is:
pred_list <- map(list("var1 = 1:10", "var2=21:30"), (x) predictions(my_model, newdata = datagrid(!!x)))
However, this does not give me a list of predictions for each of var1 and var2, across the specified ranges, with other predictors in the model held at default values.
Any help would be much appreciated.
1
You can use do.call() to feed arguments to the datagrid() function. Note that you need to specify the model object explicitly because of scoping issues:
library(marginaleffects)
mod <- lm(mpg ~ hp + wt, data = mtcars)
args <- list(list(wt = 1:3, model = mod), list(hp=100:103, model = mod))
lapply(args, function(x) predictions(mod, newdata = do.call(datagrid, x)))
> [[1]]
>
> wt Estimate Std. Error z Pr(>|z|) S 2.5 % 97.5 % hp
> 1 28.7 1.476 19.4 <0.001 277.2 25.8 31.6 147
> 2 24.8 0.896 27.7 <0.001 557.8 23.1 26.6 147
> 3 20.9 0.479 43.7 <0.001 Inf 20.0 21.9 147
>
> Columns: rowid, estimate, std.error, statistic, p.value, s.value, conf.low, conf.high, hp, wt, mpg
> Type: response
>
>
> [[2]]
>
> hp Estimate Std. Error z Pr(>|z|) S 2.5 % 97.5 % wt
> 100 21.6 0.623 34.6 <0.001 871.0 20.4 22.8 3.22
> 101 21.5 0.617 34.9 <0.001 885.5 20.3 22.8 3.22
> 102 21.5 0.611 35.2 <0.001 900.3 20.3 22.7 3.22
> 103 21.5 0.605 35.5 <0.001 915.2 20.3 22.7 3.22
>
> Columns: rowid, estimate, std.error, statistic, p.value, s.value, conf.low, conf.high, wt, hp, mpg
> Type: response