Poisson or negative-binomial likelihood for the daily counts x of a
segment, with mean given by growth_incidence(). Parameters are
optimised on the scale \((\log y_0, \log(K - y_0), \log\gamma[, \log\phi])\)
from a small grid of starting values.
Usage
fit_growth_incidence(
x,
model = c("logistic", "gompertz", "richards"),
family = c("poisson", "negbin"),
beta = 1.5,
start = NULL
)
Arguments
- x
Daily counts of the segment (local times 1, ..., length(x)).
- model
Growth model.
- family
"poisson" or "negbin".
- beta
Richards shape (held fixed).
- start
Optional warm start on the unconstrained scale (e.g. par
of a previous fit), tried in addition to the default grid.
Value
A list with coef, loglik, aic, bic, k (number of
estimated parameters), fitted, model, family, beta, n.
