Best single-model split MSE at a candidate partition point (fast proxy)

A single-growth-model analogue of split_mse(), roughly three times cheaper (one model fit per side instead of three), used internally by permutation_test_partition() to generate the null distribution of the grid-maximum MSE-reduction statistic at a cost that keeps a full permutation test tractable. The model used is null_model (default "gompertz"), fixed across the whole null simulation, not re-selected per replicate.

Usage

split_mse_single(t, y, d, null_model = "gompertz", beta_fixed = 1.5)

Arguments

t, y

Time index and cumulative counts of the (open) window being searched, both starting at 1.

d

Candidate partition point, 1 <= d < length(y).

null_model

One of "logistic", "gompertz", "richards".