multiple_lms_tidyverse.R

# quick linear regression using broom and the tidyverse ####

# load packages
library(tidyr)
library(ggplot2)
library(dplyr)
library(broom)
library(gapminder)
library(janitor)
library(purrr)

# load dataset
data(gapminder)

# clean up names using janitor::clean_names()
d <- clean_names(gapminder)

# look at data
d %>%
  glimpse()

# plot life_exp by year by continent
ggplot(d, aes(year, life_exp, col = continent)) +
  geom_point() +
  stat_smooth(method = 'lm', se = FALSE)

# lets fit an lm for each continent
d_mods <- group_by(d, continent) %>%
  nest() %>%
  mutate(., fit = purrr::map(data, ~lm(life_exp ~ year, data = .x)))

# look at first fit
d_mods$fit[[1]]

# look at model dataframe
d_mods

# get mod diagnostics - using broom::glance
mod_diags <- d_mods %>%
  unnest(fit %>% map(glance))
mod_diags

# get parameters using broom::tidy
params <- d_mods %>%
  unnest(fit %>% map(tidy))
params

# get preds using broom::augment
preds <- d_mods %>%
  unnest(fit %>% map(augment))

# plot life_exp by year by continent
ggplot(preds, aes(year, life_exp, col = continent)) +
  geom_point() +
  geom_line(aes(year, .fitted))