library(palmerpenguins)
library(dplyr)
#' @param rmse Root mean squared error on your sample
#' @param df Degrees of Freedom in your model. In this case it should be the
#' same as the number of observations in your sample.
rmse_interval <- function(rmse, deg_free, p_lower = 0.025, p_upper = 0.975){
tibble(.pred_lower = sqrt(deg_free / qchisq(p_upper, df = deg_free)) * rmse,
.pred_upper = sqrt(deg_free / qchisq(p_lower, df = deg_free)) * rmse)
}
penguins <- palmerpenguins::penguins %>%
na.omit()
split <- rsample::initial_split(penguins)
train <- rsample::training(split)
holdout <- rsample::testing(split)
mod <- lm(body_mass_g ~ bill_length_mm + bill_depth_mm + sex + species, data = train)
preds <- bind_cols(
tibble(.pred = predict(mod, holdout)),
relocate(holdout, body_mass_g)
)
# rmse example
preds %>%
summarise(rmse = yardstick::rmse_vec(body_mass_g, .pred),
n = n()) %>%
mutate(rmse_interval(rmse, n))
#> # A tibble: 1 x 4
#> rmse n .pred_lower .pred_upper
#> <dbl> <int> <dbl> <dbl>
#> 1 327. 83 284. 386.
# example just adding column for standard error#
preds %>%
summarise(n = n(),
mae = yardstick::mae_vec(body_mass_g, .pred)) %>%
mutate(mae_se = mae / sqrt(n),
mae_lower = mae - 1.96 * mae_se,
mae_upper = mae + 1.96 * mae_se)
#> # A tibble: 1 x 5
#> n mae mae_se mae_lower mae_upper
#> <int> <dbl> <dbl> <dbl> <dbl>
#> 1 83 244. 26.8 192. 297.Created on 2021-03-17 by the reprex package (v0.3.0)