Exposure metrics selection for Emax models — dev_ermod_emax_exp

This functions is used to develop an Emax model with binary and continuous endpoint, using various exposure metrics and selecting the best one.

Usage

dev_ermod_emax_exp_sel(
  data,
  var_resp,
  var_exp_candidates,
  verbosity_level = 1,
  chains = 4,
  iter = 2000,
  gamma_fix = 1,
  e0_fix = NULL,
  emax_fix = NULL,
  priors = NULL,
  seed = sample.int(.Machine$integer.max, 1)
)

dev_ermod_bin_emax_exp_sel(
  data,
  var_resp,
  var_exp_candidates,
  verbosity_level = 1,
  chains = 4,
  iter = 2000,
  gamma_fix = 1,
  e0_fix = NULL,
  emax_fix = NULL,
  priors = NULL,
  seed = sample.int(.Machine$integer.max, 1)
)

Arguments

data: Input data for E-R analysis
var_resp: Response variable name in character
var_exp_candidates: Candidate exposure variable names in character vector
verbosity_level: Verbosity level. 0: No output, 1: Display steps, 2: Display progress in each step, 3: Display MCMC sampling.
chains: Number of chains for Stan.
iter: Number of iterations for Stan.
gamma_fix: Hill coefficient, default fixed to 1. See details in rstanemax::stan_emax() or rstanemax::stan_emax_binary()
e0_fix: See details in rstanemax::stan_emax() or rstanemax::stan_emax_binary()
emax_fix: See details in rstanemax::stan_emax() or rstanemax::stan_emax_binary()
priors: See details in rstanemax::stan_emax() or rstanemax::stan_emax_binary()
seed: Random seed for Stan model execution, see details in rstan::sampling() which is used in rstanemax::stan_emax() or rstanemax::stan_emax_binary()

Value

An object of class ermod_emax_exp_sel or ermod_bin_emax_exp_sel.

Examples

# \donttest{
data_er_cont <- rstanemax::exposure.response.sample
noise <- 1 + 0.5 * stats::rnorm(length(data_er_cont$exposure))
data_er_cont$exposure2 <- data_er_cont$exposure * noise
# Replace exposure < 0 with 0
data_er_cont$exposure2[data_er_cont$exposure2 < 0] <- 0

ermod_emax_exp_sel <-
  dev_ermod_emax_exp_sel(
    data = data_er_cont,
    var_resp = "response",
    var_exp_candidates = c("exposure", "exposure2")
  )
#> ℹ The exposure metric selected was: exposure

ermod_emax_exp_sel
#> 
#> ── Emax model & exposure metric selection ──────────────────────────────────────
#> ℹ Use `plot_er_exp_sel()` for ER curve of all exposure metrics
#> ℹ Use `plot_er()` with `show_orig_data = TRUE` for ER curve of the selected exposure metric
#> 
#> ── Exposure metrics comparison ──
#> 
#>           elpd_diff se_diff
#> exposure   0.00      0.00  
#> exposure2 -9.46      3.30  
#> 
#> ── Selected model ──
#> 
#> ---- Emax model fit with rstanemax ----
#>        mean se_mean    sd  2.5%   25%   50%   75%  97.5%   n_eff Rhat
#> emax  92.01    0.12  6.07 80.13 88.01 92.02 96.20 103.86 2418.78    1
#> e0     5.72    0.10  4.72 -3.63  2.54  5.74  8.88  14.96 2085.81    1
#> ec50  75.96    0.49 20.45 45.07 61.24 73.36 87.46 124.72 1769.73    1
#> gamma  1.00     NaN  0.00  1.00  1.00  1.00  1.00   1.00     NaN  NaN
#> sigma 16.63    0.03  1.56 13.92 15.50 16.51 17.58  19.96 2735.09    1
#> * Use `extract_stanfit()` function to extract raw stanfit object
#> * Use `extract_param()` function to extract posterior draws of key parameters
#> * Use `plot()` function to visualize model fit
#> * Use `posterior_predict()` or `posterior_predict_quantile()` function to get
#>   raw predictions or make predictions on new data
#> * Use `extract_obs_mod_frame()` function to extract raw data 
#>   in a processed format (useful for plotting)
# }

# \donttest{
data_er_bin <- rstanemax::exposure.response.sample.binary

noise <- 1 + 0.5 * stats::rnorm(length(data_er_bin$conc))
data_er_bin$conc2 <- data_er_bin$conc * noise
data_er_bin$conc2[data_er_bin$conc2 < 0] <- 0

ermod_bin_emax_exp_sel <-
  dev_ermod_bin_emax_exp_sel(
    data = data_er_bin,
    var_resp = "y",
    var_exp_candidates = c("conc", "conc2")
  )
#> ℹ The exposure metric selected was: conc
# }