This function returns coefficients from a model fit by SLOPE()
.
Arguments
- object
an object of class
'SLOPE'
.- alpha
penalty parameter for SLOPE models; if
NULL
, the values used in the original fit will be used- exact
if
TRUE
and the given parameter values differ from those in the original fit, the model will be refit by callingstats::update()
on the object with the new parameters. IfFALSE
, the predicted values will be based on interpolated coefficients from the original penalty path.- simplify
if
TRUE
,base::drop()
will be called before returning the coefficients to drop extraneous dimensions- intercept
whether to include the intercept in the output; only applicable when
simplify = TRUE
and an intercept has been fit.- scale
whether to return the coefficients in the original scale or in the normalized scale.
- sigma
deprecated. Please use
alpha
instead.- ...
arguments that are passed on to
stats::update()
(and therefore also toSLOPE()
) ifexact = TRUE
and the given penalty is not inobject
Details
If exact = FALSE
and alpha
is not in object
,
then the returned coefficients will be approximated by linear interpolation.
If coefficients from another type of penalty sequence
(with a different lambda
) are required, however,
please use SLOPE()
to refit the model.
See also
Other SLOPE-methods:
deviance.SLOPE()
,
plot.SLOPE()
,
predict.SLOPE()
,
print.SLOPE()
,
score()
Examples
fit <- SLOPE(mtcars$mpg, mtcars$vs, path_length = 10)
coef(fit)
#> 2 x 8 sparse Matrix of class "dgCMatrix"
#>
#> [1,] 0.4375 -0.27721605 -0.53407168 -0.62638081 -0.65955499 -0.67147717
#> [2,] . 0.03557461 0.04835946 0.05295409 0.05460532 0.05519874
#>
#> [1,] -0.6757618 -0.67730159
#> [2,] 0.0554120 0.05548865
coef(fit, scale = "normalized")
#> 2 x 8 sparse Matrix of class "dgCMatrix"
#>
#> [1,] 0.4375 0.43750000 0.43750000 0.43750000 0.43750000 0.43750000 0.437500
#> [2,] . 0.03557461 0.04835946 0.05295409 0.05460532 0.05519874 0.055412
#>
#> [1,] 0.43750000
#> [2,] 0.05548865