Prints the backend, the formula, the number of observations and the CRS, then lines for the backend:
a random forest (
rf_fit): the number of trees,mtry, the node size, whether the coordinates were used as predictors, how each tree's rows were sampled, the out-of-bag error and the largest importances. A forest fitted without them prints "no spatial information used": it sees the covariates alone. One fitted withinclude_coords = TRUEprints a note on how to score it (see "Coordinates as predictors" and "Out-of-bag error" infit_rf_model()). When a predictor is an unordered factor or a character column, a line says how ranger was told to split it;a GWR (
gwr_fit): the bandwidth (a count of neighbours, or a distance with the CRS's unit), the kernel, the AICc, the effective number of parameters and, when any local window is of deficient rank, the number of such windows;a Bayesian GP (
bayesian_fit): the GP basis size, the smallest effective range the basis resolves in CRS units, the LOOIC when the fit computed it (compute_loo = TRUE, the default offit_bayesian_spatial_model()), the slope prior, the standardised predictors, and a flagged line when the sampler's diagnostics or the basis check failed.
A custom subclass prints the common lines only. Typing the object's name
calls this method; use it to confirm that a fit used the data, predictors
and CRS you meant. For fit quality use model_metrics() or
summary.spatial_fit(). Apart from a forest's out-of-bag
error and a Bayesian GP's LOOIC, an approximate leave-one-out score of
single rows, nothing printed here is an out-of-sample score, and neither
of those is a spatial one.
See also
fit_rf_model(), fit_gwr_model() and
fit_bayesian_spatial_model(), which make the objects.
Other fit methods:
plot.spatial_fit(),
spatial_fit-coef,
spatial_fit-fitted,
spatial_fit-predict,
spatial_fit-residuals,
summary.spatial_fit()
Examples
library(sf)
set.seed(1)
pts <- st_as_sf(
data.frame(x = 5e5 + runif(60, 0, 1000), y = 5e6 + runif(60, 0, 1000),
a = rnorm(60)),
coords = c("x", "y"), crs = 32632
)
pts$z <- 10 + 2 * pts$a + rnorm(60, 0, 0.3)
fit <- fit_rf_model(pts, "z", "a", num_trees = 50, seed = 1)
fit # print.rf_fit()
#> <Random Forest (ranger)> spatial model fit
#> Formula : z ~ a
#> n : 60
#> CRS : EPSG:32632
#> Trees : 50 (mtry = 1, min node = 5)
#> Coords as predictors: no - no spatial information used; see ?fit_rf_model
#> Sampling: bootstrap, with replacement (100.0% of rows per tree)
#> OOB RMSE: 0.4094 OOB R^2: 0.9502
#> Importance (permutation): a=6.034
#>
#> OOB is a random hold-out: under spatial autocorrelation it is the error
#> of predicting among the samples, and reads low for predictions made
#> farther from the data than the samples sit from each other. Score a map
#> that predicts farther with cv_rf() on folds at its distances
#> (?fold_separation).
if (requireNamespace("sp", quietly = TRUE))
print(fit_gwr_model(pts, "z", "a", bandwidth = 30)) # print.spatial_fit()
#> <GWR> spatial model fit
#> Formula : z ~ a
#> n : 60
#> CRS : EPSG:32632
#> Bandwidth: 30 neighbours (adaptive, bisquare kernel)
#> AICc : 47.44
#> Eff. parameters: 13.2 of 60 observations