Extract new priors from a previous fit — post2prior • covidseir

This function takes a model fit with fit_seir() and converts the posterior distribution into priors that can be fed into a second model. This allows splitting a time series into multiple blocks.

post2prior(obj, iter = seq_len(100), time_slice = max(proj$time))

Arguments

obj	A model fit with `fit_seir()`.
iter	Iterations to use went creating projections to extract the state posterior.
time_slice	When to take the state priors at. Defaults to the end of the fit.

Examples

# See the vignette 'Fitting case data in multiple blocks'
# for a complete example

cases <- c(
  0, 0, 1, 3, 1, 8, 0, 6, 5, 0, 7, 7, 18, 9, 22, 38, 53, 45, 40,
  77, 76, 48, 67, 78, 42, 66, 67, 92, 16, 70, 43, 53, 55, 53, 29,
  26, 37, 25, 45, 34, 40, 35
)
s1 <- c(rep(0.1, 13), rep(0.2, length(cases) - 13))
m <- fit_seir(
  cases,
  iter = 200,
  fit_type = "optimizing",
  samp_frac_fixed = s1
)
#> Finding the MAP estimate.
post2prior(m, iter = seq_len(10)) # just 10 for example speed
#> $R0
#> [1] 1.21978417 0.08001301
#> 
#> $i0
#> [1] 7.64153273 0.08844256
#> 
#> $e
#> [1] 0.82600448 0.04157469
#> 
#> $phi
#> [1] 1.9357203 0.3255403
#> 
#> $f
#> [1] 0.19456721 0.08779538
#> 
#> $start_decline
#> [1] 0.0 0.1
#> 
#> $end_decline
#> [1] 0.0 0.1
#> 
#> $state_0
#>     E1_frac     E2_frac      I_frac       Q_num       R_num    E1d_frac 
#>   0.6602933   0.1150118   0.2890190  15.1256445 163.2298501   0.6125361 
#>    E2d_frac     Id_frac      Qd_num      Rd_num 
#>   0.1066115   0.2673027  66.2332312 711.7988096 
#>