BayesERtools

https://github.com/genentech/bayesertools

Repository

README.Rmd

---
output: github_document
---



```{r, include = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "75%"
)

library(dplyr)
library(tidyr)
library(gt)
```


# BayesERtools 


[![R-CMD-check](https://github.com/Genentech/BayesERtools/actions/workflows/R-CMD-check.yaml/badge.svg)](https://github.com/Genentech/BayesERtools/actions/workflows/R-CMD-check.yaml)
[![CRAN status](https://www.r-pkg.org/badges/version/BayesERtools)](https://CRAN.R-project.org/package=BayesERtools)
[![downloads](https://cranlogs.r-pkg.org/badges/grand-total/BayesERtools)](https://CRAN.R-project.org/package=BayesERtools)
[![Codecov test coverage](https://codecov.io/gh/Genentech/BayesERtools/branch/main/graph/badge.svg)](https://app.codecov.io/gh/Genentech/BayesERtools?branch=main)


`BayesERtools` provides a suite of tools that facilitate
exposure-response analysis using Bayesian methods. 

- Tutorial (`BayesERbook`): https://genentech.github.io/BayesERbook/
- Package documentation: https://genentech.github.io/BayesERtools/
- GitHub repo of the package: https://github.com/genentech/BayesERtools/

## Installation

You can install the `BayesERtools` with:

``` r
install.packages('BayesERtools')
# devtools::install_github("genentech/BayesERtools") # development version
```

## Supported model types

```{r, echo = FALSE}
set.seed(1234) # Needed to stablize div id
# Need to do this to remove CSS from the outputs for it
# to work in GitLab-flavored md
remove_css <- function(x) {
  x <- gsub("", "", x)
  htmltools::HTML(x)
}

# Define the initial transposed tibble
tab_mod_raw <- tibble(
  feature = c("lin_logit", "emax_logit", "linear", "emax"),
  backend = c("`rstanarm`", "`rstanemax`", "`rstanarm`", "`rstanemax`"),
  reference =
    c(
      "https://mc-stan.org/rstanarm/reference/stan_glm.html",
      "https://yoshidk6.github.io/rstanemax/reference/stan_emax.html",
      "https://mc-stan.org/rstanarm/reference/stan_glm.html",
      "https://yoshidk6.github.io/rstanemax/reference/stan_emax_binary.html"
    ),
  `develop model` = c("✅", "✅", "✅", "✅"),
  `simulate & plot ER` = c("✅", "✅", "✅", "✅"),
  `exposure metrics selection` = c("✅", "✅", "✅", "✅"),
  `covariate selection` = c("✅", "❌", "✅", "❌"),
  `covariate forest plot` = c("✅", "❌", "✅", "❌")
)

# Transpose the table for display
tab_mod <- tab_mod_raw %>%
  pivot_longer(
    cols = -feature,
    names_to = "feature_name", values_to = "value"
  ) %>%
  pivot_wider(names_from = feature, values_from = value) |>
  mutate(.row_id = row_number())

readr::write_csv(tab_mod, "vignettes/data/supported_models.csv")

tab_mod |>
  select(-.row_id) |>
  gt() |>
  fmt_markdown() |>
  fmt_url(
    columns = !1,
    rows = 2,
    label = "🔗",
    show_underline = FALSE
  ) |>
  tab_spanner(
    label = "Binary endpoint",
    columns = c(lin_logit, emax_logit)
  ) |>
  tab_spanner(
    label = "Continuous endpoint",
    columns = c(linear, emax)
  ) |>
  cols_label(
    feature_name = "",
    lin_logit = "Linear (logit)",
    emax_logit = md("Emax (logit)"),
    linear = "Linear",
    emax = md("Emax"),
  ) |>
  tab_style(
    style = cell_text(v_align = "top", align = "center"),
    locations = cells_column_labels()
  ) |>
  tab_style(
    style = cell_text(v_align = "middle", align = "center"),
    locations = cells_body()
  ) |>
  tab_style(
    style = cell_text(v_align = "middle", align = "right"),
    locations = cells_body(columns = feature_name)
  ) |>
  tab_footnote(
    footnote = paste(
      "✅ Available",
      "🟡 In plan/under development",
      "❌ Not in a current plan",
      sep = ", "
    )
  ) |>
  as_raw_html(inline_css = FALSE) |>
  remove_css()
```


## Quick guide

Here is a quick demo on how to use this package for E-R analysis. 
See [Basic workflow](https://genentech.github.io/BayesERbook/notebook/binary/basic_workflow.html) for 
more thorough walk through.

```{r, warning=FALSE, message=FALSE}
# Load package and data
library(dplyr)
library(BayesERtools)
ggplot2::theme_set(ggplot2::theme_bw(base_size = 12))

data(d_sim_binom_cov)

# Hyperglycemia Grade 2+ (hgly2) data
df_er_ae_hgly2 <-
  d_sim_binom_cov |>
  filter(AETYPE == "hgly2") |>
  # Re-scale AUCss, baseline age
  mutate(
    AUCss_1000 = AUCss / 1000, BAGE_10 = BAGE / 10,
    Dose = paste(Dose_mg, "mg")
  )

var_resp <- "AEFLAG"
```

### Simple univariable model for binary endpoint

```{r ermod_bin, fig.width = 6, fig.height = 4.5, dpi = 150}
set.seed(1234)
ermod_bin <- dev_ermod_bin(
  data = df_er_ae_hgly2,
  var_resp = var_resp,
  var_exposure = "AUCss_1000"
)
ermod_bin

# Using `*` instead of `+` so that scale can be
# applied for both panels (main plot and boxplot)
plot_er_gof(ermod_bin, var_group = "Dose", show_coef_exp = TRUE) *
  xgxr::xgx_scale_x_log10(guide = ggplot2::guide_axis(minor.ticks = TRUE))
```

### Covariate selection

BGLUC (baseline glucose) is selected while other two covariates are not.

```{r ermod_bin_cov_sel, fig.width = 6, fig.height = 4, dpi = 150}
set.seed(1234)
ermod_bin_cov_sel <-
  dev_ermod_bin_cov_sel(
    data = df_er_ae_hgly2,
    var_resp = var_resp,
    var_exposure = "AUCss_1000",
    var_cov_candidate = c("BAGE_10", "RACE", "BGLUC")
  )
ermod_bin_cov_sel
plot_submod_performance(ermod_bin_cov_sel)
```

```{r plot_coveff, fig.width = 5, fig.height = 3, dpi = 150}
coveffsim <- sim_coveff(ermod_bin_cov_sel)
plot_coveff(coveffsim)
```

Owner

GitHub Events

Issues and Pull Requests

Last synced: 9 months ago

Packages