I’m new to Bayes and `brms`

. Apologize if this question has been asked before (a brief search did not help). I’m following the radon examples in Gelman and Hill (2007).

Suppose I fit a multilevel model in `brms`

:

```
library(brms)
f1 <- brm(log_radon ~ floor + (1 | county), data = radon, family = "gaussian")
```

I want to plot the predictions by county for each level of floor (fig 12.4 p.257 of Gelman and Hill). I suppose one way is:

```
radon_yhats <- radon %>%
mutate(log_radon_multilev = predict(f1, radon))
library(ggplot2)
ggplot(data = radon_yhats, aes(y = log_radon, x = floor)) +
geom_jitter() +
geom_line(aes(x = floor, y = log_radon_multilev)) +
facet_wrap(vars(county))
```

However, elsewhere I’ve seen references to `tidybayes::add_epred_draws()`

or sometimes people use `modlr`

.

I guess I’m a bit confused about the difference between these approaches. For instance, is `tidybayes::add_epred_draws()`

the same as `predict()`

? What would be tidy-esque approach to plotting this figure?

Many thanks.