Safe predictions from a smooth.spline object

We strongly recommend using mgcv::gam() instead of smooth.spline(). mgcv provides a feature-rich and frequently updated GAM implementation.

# S3 method for smooth.spline
safe_predict(object, new_data, type = "response",
  ...)

Arguments

object	A smooth.spline object returned from a call to stats::smooth.spline().
new_data	A numeric vector. Can contain NAs, although the presence of any NAs may slow down prediction.
type	What kind of predictions to return. Options are: "response" (default): Standard predictions from smoothing splines.
...	Unused. safe_predict() checks that all arguments in ... are evaluated via the ellipsis package. The idea is to prevent silent errors when arguments are mispelled. This feature is experimental and feedback is welcome.

Value

A tibble::tibble() with one row for each row of new_data. Predictions for observations with missing data will be NA. Returned tibble has different columns depending on type:

• "response":

  • univariate outcome: .pred (numeric)

  • multivariate outcomes: .pred_{outcome name} (numeric) for each outcome

• "class": .pred_class (factor)

• "prob": .pred_{level} columns (numerics between 0 and 1)

• "link": .pred (numeric)

• "conf_int": .pred, .pred_lower, .pred_upper (all numeric)

• "pred_int": .pred, .pred_lower, .pred_upper (all numeric)

If you request standard errors with std_error = TRUE, an additional column .std_error.

For interval predictions, the tibble has additional attributes level and interval. The level is the same as the level argument and is between 0 and 1. interval is either "confidence" or "prediction". Some models may also set a method attribute to detail the method used to calculate the intervals.

Uncertainty in predictions

If you use mgcv, mgcv::predict.gam() has a se.fit argument that will allow you get standard errors for predictions.

If you insist on using smooth.spline(), your best bet to estimate uncertainty in predictions is bootstrapping. See

vignette("bootstrapping", package = "safepredict")

for worked examples and details on how you might do this.

Derivatives of smooths

While stats:::predict.smooth.spline() has a deriv argument we do not support it. If you would like to estimate derivates of smooths we recommend using the gratia package, and in particular gratia::fderiv(). At the time of writing, gratia is not yet on CRAN, but can be installed from Github with devtools::install_github(gavinsimpson/gratia). Additional documentation on gratia is available here.

Examples

fit <- smooth.spline(mtcars$mpg, mtcars$wt, cv = TRUE)
#> Warning: cross-validation with non-unique 'x' values seems doubtful
safe_predict(fit, c(30, NA, 40))
#> # A tibble: 3 x 1
#>   .pred
#>   <dbl>
#> 1  1.91
#> 2 NA   
#> 3  1.06

# the following will fail, however
# NOT RUN {
predict(fit, c(30, NA, 40))
# }

# however, we recommend using mgcv instead
library(mgcv)
#> Loading required package: nlme
#> This is mgcv 1.8-23. For overview type 'help("mgcv-package")'.

fit2 <- gam(mpg ~ s(wt), data = mtcars)
predict(fit2, mtcars)  # TODO: update with safe_predict once implemented
#>           Mazda RX4       Mazda RX4 Wag          Datsun 710      Hornet 4 Drive 
#>            23.03365            21.30454            25.24007            19.21380 
#>   Hornet Sportabout             Valiant          Duster 360           Merc 240D 
#>            17.95427            17.84824            17.28478            19.36025 
#>            Merc 230            Merc 280           Merc 280C          Merc 450SE 
#>            19.59685            17.95427            17.95427            15.11330 
#>          Merc 450SL         Merc 450SLC  Cadillac Fleetwood Lincoln Continental 
#>            16.52446            16.30057            11.46772            11.00221 
#>   Chrysler Imperial            Fiat 128         Honda Civic      Toyota Corolla 
#>            11.21332            26.15766            30.67862            28.98223 
#>       Toyota Corona    Dodge Challenger         AMC Javelin          Camaro Z28 
#>            24.15407            17.53673            17.98094            16.03997 
#>    Pontiac Firebird           Fiat X1-9       Porsche 914-2        Lotus Europa 
#>            16.01864            28.20817            26.62028            31.46338 
#>      Ford Pantera L        Ferrari Dino       Maserati Bora          Volvo 142E 
#>            19.47820            21.99848            17.28478            21.93130 

# to get estimated derivatives from smooths use gratia,
# which extends mgcv

# NOT RUN {
gratia::fderiv(fit2)  # TODO: gratia isn't on CRAN yet
# }