Returns P(LOS > d) for d = 0, 1, ..., max_stay. Use this when
you do not have observed census data to fit on and want to plug in
literature priors. The returned numeric vector drops straight into
fcast_census(). Counterpart of fit_los(), which recovers the
same vector from data.
Usage
spec_los(family, ..., max_stay = 50)
Arguments
- family
Character. One of "negbin", "normal",
"lognormal", "geometric".
- ...
Parameters for the chosen family (see above).
- max_stay
Integer. Longest LOS modelled, in time step units
of the forecast (weeks for FluSight, days for daily hubs). The
returned vector has length max_stay + 1.
Value
Numeric vector of length max_stay + 1 giving
P(LOS > d) for d = 0..max_stay.
Details
Parameters per family:
negbin: mu (mean), k (dispersion).
normal: mu (mean), sigma (sd).
lognormal: meanlog, sdlog.
geometric: mu (mean).
Examples
spec_los("negbin", mu = 2, k = 1.7, max_stay = 15)
#> [1] 0.7334244982 0.4884632262 0.3097077035 0.1905373549 0.1148480795
#> [6] 0.0682071206 0.0400544698 0.0233150557 0.0134749948 0.0077423467
#> [11] 0.0044267070 0.0025204178 0.0014298830 0.0008086636 0.0004560796
#> [16] 0.0002565996
