update.earth {earth} | R Documentation |
Update an ‘earth’ model.
## S3 method for class 'earth': update(object = stop("no 'object' arg"), formula. = NULL, ponly = FALSE, ..., evaluate = TRUE)
object |
|
formula. |
These arguments are the same as earth .
The formula. argument is treated like earth's formula argument. |
ponly |
Force only pruning, no forward pass. Default is FALSE, meaning decide automatically if a forward pass is needed. See note below. |
... |
Arguments passed on to earth .
|
evaluate |
If TRUE (default) evaluate the new call else return the call.
Mostly for compatibility with the generic update .
|
The value is the same as that returned by earth
.
If object
is the only parameter then no changes are made
— the returned value will be the same as the original object
.
If only the following arguments are used, a forward pass
is unnecessary, and update.earth
will perform only the pruning pass.
This is usually much faster for large models.
trace pmethod nprune Get.crit Eval.model.subsets Print.pruning.pass Force.xtx.prune Use.beta.cacheThis automatic determination to do a forward pass can be overriden with the
ponly
argument.
If ponly=TRUE
the forward pass will be skipped, and only the pruning pass will be executed.
This is useful for doing a pruning pass with new data, typically with penalty=1
.
A similar use of ponly=TRUE
is to do subset
selection with a different penalty
from that used to build the original model.
data(ozone1) (a <- earth(O3 ~ ., data = ozone1, degree = 2)) # yields: # Selected 11 of 21 terms, and 8 of 9 predictors # Number of terms at each degree of interaction: 1 5 5 # GCV: 13.43422 RSS: 3762.234 GRSq: 0.7913151 RSq: 0.8218252 update(a, formula = O3 ~ . - temp) # requires forward pass and pruning # yields: # Selected 15 of 21 terms, and 8 of 8 predictors # Number of terms at each degree of interaction: 1 5 9 # GCV: 13.14588 RSS: 3443.265 GRSq: 0.7957941 RSq: 0.8369311 update(a, nprune = 8) # requires only pruning # yields: # Selected 8 of 21 terms, and 6 of 9 predictors # Number of terms at each degree of interaction: 1 5 2 # GCV: 15.07944 RSS: 4433.915 GRSq: 0.7657586 RSq: 0.7900152 update(a, penalty=1, ponly=TRUE) # pruning pass only with a new penalty # yields: # Selected 13 of 21 terms, and 8 of 9 predictors # Number of terms at each degree of interaction: 1 5 7 # GCV: 12.5 RSS: 3675 GRSq: 0.805 RSq: 0.826