
Print, summarize and predict from a fitted emulator
Source:R/emulator_methods.R
spDBL_emulator-methods.RdMethods for the result of emulator_learn(). print() shows the model
settings. summary() adds the in-sample fit of the smoothed coefficients
and the posterior mean of the residual variance. predict() calls
emulator_predict(), or with interval = "prediction" returns pointwise
predictive distributions.
Arguments
- x, object
An
spDBL_emulatorfromemulator_learn().- ...
Unused.
- newdata
Matrix (or vector, for one run) of new mechanistic inputs.
- dt_pde_test, F_new_ls, MC
Passed to
emulator_predict().- interval
"none"(the default) returns whatemulator_predict()returns."prediction"returns the pointwise predictive distribution given the autoregressive rows: for each site and time, a Student-t distribution from integrating the coefficients, the residual covariance and the input Gaussian process analytically.- level
Coverage of the central prediction intervals.
Value
print() returns x invisibly. summary() returns an object of
class summary.spDBL_emulator, a list with
- settings
One-row data frame of model settings.
- rmse_by_time
Root mean squared error of the smoothed fit \(F_e \hat\Theta_e\) to the training responses, at each time.
- rmse
The same over all times.
- response_sd
Standard deviation of the training responses.
- residual_variance
Posterior mean of the residual variance, averaged over the sites of a block: \(D/(n-C-1)\) for the inverse-Wishart model and \(D/(n-1)\) times the mean diagonal of
R_episodefor the \(\sigma^2 R\) model, from the final filter.
predict() returns what emulator_predict() returns, or, with
interval = "prediction", a list with time-indexed lists mean, sd,
lower and upper of new-run by site matrices (the bounds of central
intervals with coverage level), plus level, distribution and the
degrees of freedom df.
Examples
data("dt_emulation")
d <- dt_emulation
em <- emulator_learn(d$pde_para_train, d$dt_pde_train, Nx = 10, Ny = 10,
nsam = 5, seed = 1)
em
#> spDBL emulator (inverse-Wishart covariance, marginal coefficient sampling)
#> Training runs: 40 with 5 inputs
#> Grid: 10 x 10 (100 sites) over 26 times
#> Episodes: 260 = 26 times x 10 blocks of 10 sites
#> Design: AR(2), 20 columns
#> Stored posterior draws: 5
summary(em)
#> spDBL emulator summary
#> Covariance: iw; sampling: marginal; AR order: 2
#> 40 runs x 100 sites x 26 times; 260 episodes; 5 stored draws
#> In-sample RMSE of the smoothed fit: 0.05785 (response SD 3.137)
#> Posterior mean residual variance per site: 0.003574
pred <- predict(em, d$pde_para_test, dt_pde_test = d$dt_pde_test)
dim(pred[[3]])
#> [1] 10 100
band <- predict(em, d$pde_para_test, dt_pde_test = d$dt_pde_test,
interval = "prediction")
# Share of held-out outputs at time 3 inside the 95% prediction intervals
mean(d$dt_pde_test[[3]] >= band$lower[[3]] & d$dt_pde_test[[3]] <= band$upper[[3]])
#> [1] 0.918