add_ci.sm_regression: Add bootstrapped model estimates

In ddsjoberg/sjosmooth: Kernel smoothing for time to event data

Description

This function is meant to be run after 'sm_regression' and will calculate bootstraped model estimates.

Usage

## S3 method for class 'sm_regression'
add_ci(x, n = 200, seed = NULL, ...)

Arguments

x	'sm_regression' object
n	number of bootstrap models to run. Default is 200.
seed	a single value, interpreted as an integer. See set.seed
...	further arguments passed to or from other methods.

Examples

sm_regression(
  data = mtcars,
  method = "lm",
  formula = mpg ~ am ,
  weighting_var = "hp",
  lambda = 1,
  newdata = data.frame(hp = c(150, 200))
) %>%
add_ci(n = 50)

# example plotting slope coefficient with confidence intervals
## Not run: 
library(ggplot2)
mtcars %>%
  # calculate locally-weighted regression models
  sm_regression(
    method = "lm",
    formula = mpg ~ am ,
    weighting_var = "hp",
    newdata = data.frame(hp = seq(125, 175, by = 5))
  ) %>%
  # add models from bootstrap resamples
  add_ci(n = 200, seed = 23948) %>%
  # calculating confidence interval for beta coefficient
  dplyr::mutate(
    # extracting central estimate of beta
    .coef = purrr::map_dbl(
      .model,
      ~ .x %>% coef() %>% purrr::pluck("am")
    ),
    # extracting each estimate of beta from bootstrapped models
    .coef.boot = purrr::map(
      .model.boot,
      ~purrr::map_dbl(
        .x,
        ~ .x %>% coef() %>% purrr::pluck("am")
      )
    ),
    # calculating the SD of the beta distribution
    .coef.sd = purrr::map_dbl(
      .coef.boot,
      sd,
      na.rm = TRUE
    ),
    # calculating the confidence interval for beta coef
    .coef.ll = .coef - qnorm(0.975) * .coef.sd,
    .coef.ul = .coef + qnorm(0.975) * .coef.sd,
  ) %>%
  ggplot(aes(x = hp, y = .coef)) +
    geom_line() +
    geom_ribbon(
      aes(ymin = .coef.ll, ymax = .coef.ul),
      alpha = 0.4
    ) +
    labs(
      y = "Slope coefficient for 'am' when regressed on 'mpg'"
    )

## End(Not run)

ddsjoberg/sjosmooth documentation built on May 14, 2019, 5:16 p.m.