Wrangles inputs into the format required for run_stan_interfaces()

Wrangles inputs into the format required for run_stan_interfaces()

prepare_data_for_stan(
  df_sam,
  df_cal,
  df_priors_scalars,
  df_priors_vectors,
  rho_prior_eta = 1,
  x_mix_pred_vars_names = list(p_pos = character(), mu_pos = character(), sd_pos =
    character(), mu_neg = character(), sd_neg = character()),
  p_pos_binary_pred_vars = character(),
  f_pred_vars_names = character()
)

Arguments

df_sam

a dataframe of the format output by simulate_data() (inside its list of outputs): with one row per sample replicate, including columns id_sam, y, and plate. y should be numeric.

df_cal

a dataframe of the format output by simulate_data() (inside its list of outputs): with one row per calibrator replicate, including columns x, y, and plate. x and y should be numeric.

df_priors_scalars, df_priors_vectors

dataframes of the format output by read_priors() (inside its list of outputs): each having one row per parameter, including columns param, lower, and upper. param should be character, lower and upper should be numeric.

rho_prior_eta

the single positive number 'eta' that is a hyperparameter of the LKJ distribution used as the prior for the correlation matrix for variability in the parameters of the x<->y relationship between plates. See the Stan docs on the LKJ distribution for the definition of eta.

x_mix_pred_vars_names

a list (whose names are the parameters p_pos, mu_neg, mu_pos, sd_neg, sd_pos) of character vectors (each one naming the categorical predictor variables used in the regression model for that parameter). The default value of a list with all empty character vectors is appropriate for using no regression model for any of these five parameters.

p_pos_binary_pred_vars

a character vector containing the names of binary variables used as fixed effects in the regression mode for seroprevalence. The default value of an empty character vector is appropriate for using no such regression model.

f_pred_vars_names

a character vector containing the names of variables used for a regression model for the four vector f. The default value of an empty character vector is appropriate for using no regression model.

Value

a list whose elements are

• df_plate, a dataframe with one row per plate;

• data_descriptors, a list of things required as input for rename_params_from_stan() and wrangle_true_params();

• stan_input_posterior, a list of things required as input for run_stan_interfaces() if you want to sample from the posterior;

• stan_input_prior, a list of things required as input for run_stan_interfaces() if you want to sample from the prior (this reduces the number of samples to zero but keeps the same 'shape' of the dataset in terms of covariates used, to more efficiently sample population-level parameters while avoiding sampling individual-level parameters such as x);

• df_sam and df_cal (as provided as input but with extra columns added for Stan indexing).