lambda(t) = exp(f(t)) with f a Gaussian process. The exponential link
is what allows an unconstrained smooth prior on f to induce a
positive hazard, and once time is binned the likelihood is exactly
Poisson with the bin exposure as offset. Ridge-smoothed bin estimates
recover a constant true hazard.
Usage
Ghosalsmhazgp(n = 500, seed = 42)
Arguments
- n
Sample size.
- seed
Seed for the deterministic draws.
Value
List with estimate (mean absolute hazard error),
hazard_by_bin, method.
Details
Formula: f_b = log((d_b + 1/2) / (E_b + 1/2)) with d_b the events and
E_b the exposure in bin b; hazard = exp(f_b).
References
Ghosal & van der Vaart (2017), Fundamentals of
Nonparametric Bayesian Inference, CUP, section 13.5.
Examples
Ghosalsmhazgp()
#> $estimate
#> [1] 0.07919604
#>
#> $hazard_by_bin
#> [1] 1.1144699 1.0054790 1.0273066 0.9532614 0.8774626 1.1586449
#>
#> $method
#> [1] "GP smooth hazard (GvdV 2017 sec. 13.5)"
#>
#> attr(,"class")
#> [1] "morie_rich_result" "list"
