neighbayes.models.priors.PanelHurdlePriors¶
-
class neighbayes.models.priors.PanelHurdlePriors(beta_mu=
None, beta_sigma=None, sigma2_alpha=2.0, sigma2_beta=None, nu=4.0, rho_lower=-1.0, rho_upper=1.0, alpha_sigma=2.5, alpha_nu=3.0, alpha_fixed=None, time_effect_mu=None, time_effect_sigma=2.5, unit_effect_mu=None, unit_effect_sd_scale=1.0, beta_mu_zero=None, beta_sigma_zero=None, rho_lower_zero=None, rho_upper_zero=None, unit_effect_mu_zero=None, unit_effect_sd_scale_zero=1.0, time_effect_mu_zero=None, time_effect_sigma_zero=2.5)[source]¶ Priors for the hurdle panels (
neighbayes.models.SARHurdleNegBin& co).The count part takes the priors of
PanelCountPriors(β, ρ, α, the unit and period effects), centred on the positive counts: the intercept defaults tolog mean(y | y > 0). The zero part’s are named as the count part’s with a_zerosuffix:- beta_mu_zero, beta_sigma_zero
Normal prior on the zero part’s coefficients; default Gelman et al. (2008) on the logit scale, the intercept centred at
logit(mean(y > 0)).
- rho_lower_zero, rho_upper_zero
Uniform bounds on the zero part’s ρ; default the count part’s.
- unit_effect_mu_zero, unit_effect_sd_scale_zero
Mean of the zero part’s unit effects (0 when its design keeps an intercept, else
logit(mean(y > 0))) and the scale of the half-t(3) prior on their learned sd (default 1).
- time_effect_mu_zero, time_effect_sigma_zero
Normal prior on each of the zero part’s period effects: mean zero beside unit effects,
logit(mean(y > 0))without them.
-
__init__(beta_mu=
None, beta_sigma=None, sigma2_alpha=2.0, sigma2_beta=None, nu=4.0, rho_lower=-1.0, rho_upper=1.0, alpha_sigma=2.5, alpha_nu=3.0, alpha_fixed=None, time_effect_mu=None, time_effect_sigma=2.5, unit_effect_mu=None, unit_effect_sd_scale=1.0, beta_mu_zero=None, beta_sigma_zero=None, rho_lower_zero=None, rho_upper_zero=None, unit_effect_mu_zero=None, unit_effect_sd_scale_zero=1.0, time_effect_mu_zero=None, time_effect_sigma_zero=2.5)[source]¶
Methods
__init__([beta_mu, beta_sigma, ...])Attributes