Extracting results

It is easy to get access to the output

Example: grab draws from the posterior for math

posterior_samples(attendance_brms, pars = 'math') %>% head()

        b_math
1 -0.005945358
2 -0.006030774
3 -0.009060251
4 -0.007864525
5 -0.006196730
6 -0.006510296

posterior_samples(attendance_brms, pars = 'math') %>% 
  qplot(data=., x = b_math, geom = 'density')

Tidy methods for data extraction

The broom package can make your model results easier to work with⁵

Convert your results to a tidy data frame and go from there!

library(broom)
tidy(attendance_brms)

            term      estimate    std.error         lower         upper
1    b_Intercept  1.486260e+00 0.0807602243  1.353464e+00  1.613702e+00
2         b_math -6.947692e-03 0.0009433085 -8.589656e-03 -5.444662e-03
3   b_genderMale -2.450094e-01 0.0444296384 -3.257716e-01 -1.680750e-01
4  b_progGeneral  1.273018e+00 0.0809187810  1.148042e+00  1.398220e+00
5 b_progAcademic  8.458137e-01 0.0716282175  7.205499e-01  9.549422e-01
6           lp__ -1.320972e+03 1.6287393824 -1.324051e+03 -1.319034e+03

library(kableExtra)
tidy(attendance_brms) %>% 
  filter(grepl(term, pattern = '^b')) %>% 
  mutate(term = c('Intercept', 'Math', 'Male', 'General', 'Academic')) %>% 
  rename_all(str_to_title) %>% 
  kable(digits = 2)

Term	Estimate	Std.error	Lower	Upper
Intercept	1.49	0.08	1.35	1.61
Math	-0.01	0.00	-0.01	-0.01
Male	-0.25	0.04	-0.33	-0.17
General	1.27	0.08	1.15	1.40
Academic	0.85	0.07	0.72	0.95

tidybayes

Bayesian analysis + tidy data + geoms

# add fitted values given the posterior draws for model parameters

library(tidybayes)
attendance %>% 
  add_fitted_draws(attendance_brms)

# A tibble: 125,600 x 10
# Groups:   id, gender, math, daysabs, prog, .row [314]
   id    gender  math daysabs prog      .row .chain .iteration .draw .value
   <fct> <fct>  <dbl>   <dbl> <fct>    <int>  <int>      <int> <int>  <dbl>
 1 1001  Male      63       4 Academic     1     NA         NA     1   5.05
 2 1001  Male      63       4 Academic     1     NA         NA     2   5.52
 3 1001  Male      63       4 Academic     1     NA         NA     3   4.83
 4 1001  Male      63       4 Academic     1     NA         NA     4   4.97
 5 1001  Male      63       4 Academic     1     NA         NA     5   5.41
 6 1001  Male      63       4 Academic     1     NA         NA     6   4.98
 7 1001  Male      63       4 Academic     1     NA         NA     7   5.37
 8 1001  Male      63       4 Academic     1     NA         NA     8   5.30
 9 1001  Male      63       4 Academic     1     NA         NA     9   4.99
10 1001  Male      63       4 Academic     1     NA         NA    10   5.20
# ... with 125,590 more rows

sleepstudy %>%
  modelr::data_grid(Days = Days,
                    Subject = levels(Subject)) %>% 
  add_predicted_draws(sleepstudy_brms) %>%
  ggplot(aes(x = Days)) +
  stat_lineribbon(aes(y = .prediction), 
                  color = NineteenEightyR::electronic_night()[1],
                  .width = seq(.5, .99, by = .01), 
                  alpha = .5,
                  show.legend = F) +
  geom_point(aes(y = Reaction), data = sleepstudy, alpha=.25) +
  scico::scale_fill_scico_d(alpha=.1, palette = 'acton', direction = -1)

Questions about tidybayes may be shouted across the street 😀

• Developed by Matthew Kay Assistant Professor at UMSI

---

5. Note that broom works with dozens of modeling packages, not just in this context.↩