Suggest model weights from fit statistics — model

Calculates transparent candidate weights from the model-fit statistics in model_table(). These weights are intended as a starting point for sensitivity analysis, not as a replacement for judgement about source data relevance.

Usage

model_weights(
  data_name = NULL,
  country = NULL,
  data_type = NULL,
  aggregation = NULL,
  mobility_model = NULL,
  model_type = NULL,
  metric = c("DIC", "RMSE", "MAPE", "R2")
)

Arguments

data_name: Optional mobility dataset name.
country: Optional ISO3C country code or vector of codes.
data_type: Optional mobility data type, such as "facebook" or "call data record".
aggregation: Optional spatial aggregation level, such as "admin_2".
mobility_model: Optional mobility model family, such as "gravity" or "departure-diffusion".
model_type: Optional model subtype, such as "exp".
metric: Fit statistic used for weighting. DIC, RMSE, and MAPE treat lower values as better. R2 treats higher values as better.

Value

A tibble::tibble() with one row per model and normalised weight.