powerscale-sensitivity: Power-scaling sensitivity analysis

In n-kall/priorsense: Prior Diagnostics and Sensitivity Analysis

Power-scaling sensitivity analysis

Description

Calculates the prior/likelihood sensitivity based on power-scaling perturbations. This is done using importance sampling (and optionally moment matching).

Usage

powerscale_sensitivity(x, ...)

## Default S3 method:
powerscale_sensitivity(
  x,
  variable = NULL,
  lower_alpha = 0.99,
  upper_alpha = 1.01,
  div_measure = "cjs_dist",
  measure_args = list(),
  component = c("prior", "likelihood"),
  sensitivity_threshold = 0.05,
  moment_match = FALSE,
  k_threshold = 0.5,
  resample = FALSE,
  transform = NULL,
  prediction = NULL,
  prior_selection = NULL,
  likelihood_selection = NULL,
  num_args = NULL,
  ...
)

## S3 method for class 'priorsense_data'
powerscale_sensitivity(
  x,
  variable = NULL,
  lower_alpha = 0.99,
  upper_alpha = 1.01,
  div_measure = "cjs_dist",
  measure_args = list(),
  component = c("prior", "likelihood"),
  sensitivity_threshold = 0.05,
  moment_match = FALSE,
  k_threshold = 0.5,
  resample = FALSE,
  transform = NULL,
  prediction = NULL,
  prior_selection = NULL,
  likelihood_selection = NULL,
  num_args = NULL,
  ...
)

## S3 method for class 'CmdStanFit'
powerscale_sensitivity(x, ...)

## S3 method for class 'stanfit'
powerscale_sensitivity(x, ...)

Arguments

x	Model fit object or priorsense_data object.
...	Further arguments passed to functions.
variable	Character vector of variables to check.
lower_alpha	Lower alpha value for gradient calculation.
upper_alpha	Upper alpha value for gradient calculation.
div_measure	The divergence measure to use. The following methods are implemented: "cjs_dist": Cumulative Jensen-Shannon distance. Default method. See function cjs_dist for more details. "js_dist": Jensen-Shannon distance. "js_div": Jensen-Shannon divergence. "hellinger_dist": Hellinger distance. "kl_dist": Kullback-Leibler distance. "kl_div": Kullback-Leibler divergence. "ks_dist": Kolmogorov-Smirnov distance. "hellinger_dist": Hellinger distance. "ws_dist": Wassterstein distance (pass measure_args = list(p = N)) for a different order, where N is the order.
measure_args	Named list of further arguments passed to divergence measure functions.
component	Character vector specifying component(s) to scale (default is both "prior" and "likelihood").
sensitivity_threshold	Threshold for flagging variable as sensitive to power-scaling.
moment_match	Logical; Indicate whether or not moment matching should be performed. Can only be TRUE if is_method is "psis".
k_threshold	Threshold value for Pareto k values above which the moment matching algorithm is used. Default is 0.5.
resample	Logical; Indicate whether or not draws should be resampled based on calculated importance weights.
transform	Indicate a transformation of posterior draws to perform before sensitivity analysis. Either "scale" or "whiten".
prediction	Function taking the model fit and returning a draws_df of predictions to be appended to the posterior draws
prior_selection	Numeric vector of prior partitions to include in power-scaling. Default is NULL, which takes all partitions.
likelihood_selection	Numeric vector of likelihood partitions to include in power-scaling. Default is NULL, which takes all partitions.
num_args	(named list) Optional arguments passed to num() for pretty printing of summaries. Can be controlled globally via the posterior.num_args option.

Value

Table of sensitivity values for each specified variable.

References

Kallioinen, N., Paananen, T., Bürkner, P-C., Vehtari, A. (2023). Detecting and diagnosing prior and likelihood sensitivity with power-scaling perturbations. Statistics and Computing. 34(57). doi:10.1007/s11222-023-10366-5

Vehtari, A., Simpson, D., Gelman, A., Yao, Y., and Gabry, J. (2024). Pareto smoothed importance sampling. Journal of Machine Learning Research. 25(72). https://jmlr.org/papers/v25/19-556.html

Paananen, T., Piironen, J., Bürkner, P-C., Vehtari, A. (2021). Implicitly adaptive importance sampling. Statistics and Computing. 31(16). doi:10.1007/s11222-020-09982-2

Examples

ex <- example_powerscale_model()
powerscale_sensitivity(ex$draws)

n-kall/priorsense documentation built on Nov. 4, 2024, 10:30 p.m.