pri_par_epsilon_grid: Grid computation for epsilon-local sensitivity

Description Usage Arguments Details Value References See Also Examples

View source: R/pri_par_epsilon_grid.R

Description

Starting with a base prior from a one-parameter family, this function determines the parameters of two priors from the same parametric family which have the given Hellinger distance to the base prior. The available one-parameter distributions are half-normal, exponential, half-Cauchy and Lomax with shape parameter fixed at 1.

Usage

1
2
pri_par_epsilon_grid(AA0_HN, AA0_EXP, AA0_HC, AA0_LMX, 
                     grid_epsilon=0.00354)

Arguments

AA0_HN

scale parameter of the base half-normal prior

AA0_EXP

scale parameter of the base exponential prior

AA0_HC

scale parameter of the base half-Cauchy prior

AA0_LMX

scale parameter of the base Lomax prior with shape parameter=1

grid_epsilon

value for epsilon, the Hellinger distance between the base prior and the
"lower"/"upper" prior on the grid

Details

This 1-dimensional epsilon grid is needed for computation of the epsilon-local sensitivity in the function sensitivity_learning_table. See Ott et al. (2020) for a brief description of this methodology and Roos et al. (2015) for more details. The default value for grid_epsilon corresponds to the Hellinger distance between two normal distribution with unit variance and a difference in means of 0.01, see Roos et al. (2015, Sect. 2.2) for calibration and interpretation of Hellinger distance values.

Value

A list of the following eight scale parameter values:

p_HN_l

parameter of the "lower" half-normal prior on the grid

p_HN_u

parameter of the "upper" half-normal prior on the grid

p_EXP_l

parameter of the "lower" exponential prior on the grid

p_EXP_u

parameter of the "upper" exponential prior on the grid

p_HC_l

parameter of the "lower" half-Cauchy prior on the grid

p_HC_u

parameter of the "upper" half-Cauchy prior on the grid

p_LMX_l

scale parameter of the "lower" Lomax prior (with shape parameter=1) on the grid

p_LMX_u

scale parameter of the "upper" Lomax prior (with shape parameter=1) on the grid

References

Ott, M., Hunanyan, S., Held, L., Roos, M. Sensitivity quantification in Bayesian meta-analysis. Manuscript revised for Research Synthesis Methods. 2020.

Roos, M., Martins, T., Held, L., Rue, H. (2015). Sensitivity analysis for Bayesian hierarchical models. Bayesian Analysis 10(2), 321–349. https://projecteuclid.org/euclid.ba/1422884977

See Also

sensitivity_learning_table, cal_h_dist

Examples

1
2
3
4
5
6
# epsilon grid for 5 % tail-adjusted priors with 
# reference threshold UU=1 and epsilon = 0.00354 (default)
pri_par_epsilon_grid(AA0_HN=pri_par_adjust_static(UU=1)$p_HN, 
                     AA0_EXP=pri_par_adjust_static(UU=1)$p_EXP,
                     AA0_HC=pri_par_adjust_static(UU=1)$p_HC,
                     AA0_LMX=pri_par_adjust_static(UU=1)$p_LMX)

sl4bayesmeta documentation built on Feb. 18, 2020, 3:02 p.m.