## kernels.simulate_one_year_cdr()


Actuary in the box: the one-year CDR distribution by re-reserving.


Usage

``` python
kernels.simulate_one_year_cdr(
    fit,
    *,
    n_draws=None,
    seed=None,
    generator=None,
    process=None,
    parameter_risk=None
)
```


One draw is one possible next year - `DiagonalGenerator.draw()` says what emerged, [rereserve()](kernels.rereserve.md#ibnr.kernels.rereserve) re-runs the volume-weighted chain ladder on the extended triangle and differences the ultimates. Returns per-origin CDR draws plus a `total` column derived from the same draws, so the diversification between accident years is in the samples; a positive draw is a reserve release.

`generator` picks the first axis: a method name from [cdr_methods()](kernels.cdr_methods.md#ibnr.kernels.cdr_methods), or a configured instance:

``` python
simulate_one_year_cdr(fit)                                  # Mack, defaults
simulate_one_year_cdr(fit, generator="odp_bootstrap")
simulate_one_year_cdr(fit, generator=ODPBootstrapDiagonal(process="gamma"))
```

`n_draws=None` means "this generator's own count", which is `DEFAULT_N_DRAWS` for the two simulating generators and the fitted posterior's own size for [ibnr.gallery.GalleryDiagonal](gallery.GalleryDiagonal.md#ibnr.gallery.GalleryDiagonal) - see `DiagonalGenerator.resolve_n_draws()`. It replaced a literal `20_000` default in 0.5.1 and changes no number: both simulating generators resolve `None` to exactly that.

`process` and `parameter_risk` are [MackDiagonal](kernels.MackDiagonal.md#ibnr.kernels.MackDiagonal)'s knobs, kept on this signature because they are the published 0.5.0 API - `None` means "that generator's default", so the defaults and every explicit call are unchanged. They cannot be combined with `generator=`; pass them to the generator instead, and see `_resolve_generator()` for why.

`seed` is anything `np.random.default_rng` accepts. The `mack` gallery entry hands a per-cohort `SeedSequence` through here; a plain integer keeps the byte-exact meaning it has always had.

`generator="merz_wuthrich"` is refused by name: the closed form is a linearization of the chain-ladder factor update around Mack's moments, not a way of generating a diagonal, and [one_year_cdr()](kernels.one_year_cdr.md#ibnr.kernels.one_year_cdr) is where it lives. `generator="gallery"` is refused too, for the opposite reason - it IS a generator, but it wraps a fitted entry that no string can carry.