Given lm object, extract the high level variable, the levels, and the reference level, as well as the estimate of the impact for this variable (assuming using effects coding).

extract_coefs_effects_coding.md

library(tidyverse)
library(broom)
library(car)

param <- getOption("contrasts")
go_deviance <- param
# traditional `contr.sum` does not name levels, so use function from `car` package
go_deviance["unordered"] <- "contr.Sum"

options(contrasts = go_deviance)

tidy_with_ref <- function(mod_obj, sep_regex = "\\[S\\."){
  
  tidy_coefs <- tidy(mod_obj) %>% 
    separate(term, c("variable", "level"), sep_regex, remove = FALSE) %>% 
    mutate(level = str_sub(level, 1, -2))
  
  xlevels <- mod_obj$xlevels  
  
  missing_levels <- xlevels %>% 
    enframe() %>% 
    unnest() %>% 
    set_names(c("variable", "level"))
  
  missing_value <- tidy_coefs %>% 
    filter(!is.na(level)) %>% 
    group_by(variable) %>% 
    summarise(estimate = sum(-estimate))
  
  missing_levels_values <- left_join(missing_levels, missing_value) %>% 
    anti_join(tidy_coefs, by = c("variable", "level"))
  
  bind_rows(tidy_coefs, missing_levels_values) %>% 
    arrange(variable, level)
}

mod_fit <- glm(cty ~ manufacturer + hwy + class, data = mpg)

tidy_with_ref(mod_fit)
#> Warning: Expected 2 pieces. Missing pieces filled with `NA` in 2 rows [1,
#> 16].
#> Joining, by = "variable"
#> # A tibble: 24 x 7
#>    term           variable   level   estimate std.error statistic   p.value
#>    <chr>          <chr>      <chr>      <dbl>     <dbl>     <dbl>     <dbl>
#>  1 (Intercept)    (Intercep~ <NA>    -0.813      0.535   -1.52     1.30e- 1
#>  2 class[S.2seat~ class      2seater -2.08       0.435   -4.77     3.47e- 6
#>  3 class[S.compa~ class      compact  0.00201    0.214    0.00939  9.93e- 1
#>  4 class[S.midsi~ class      midsize -0.681      0.199   -3.42     7.47e- 4
#>  5 class[S.miniv~ class      minivan  0.475      0.327    1.45     1.47e- 1
#>  6 class[S.picku~ class      pickup   1.47       0.236    6.21     2.81e- 9
#>  7 class[S.subco~ class      subcom~ -0.0172     0.224   -0.0767   9.39e- 1
#>  8 <NA>           class      suv      0.830     NA       NA       NA       
#>  9 hwy            hwy        <NA>     0.738      0.0217  34.0      1.00e-87
#> 10 manufacturer[~ manufactu~ audi    -0.972      0.272   -3.57     4.46e- 4
#> # ... with 14 more rows

^{Created on 2019-02-28 by the reprex package (v0.2.1)}