masher_soft: Soft masher
In bcjaeger/ipa: Imputation for Predictive Analytics

Description Usage Arguments Value Examples

If you use Rstudio, the masher and spicer functions can help remind you which parameters go along with which ipa_brew flavor. The basic idea is to write spice(brew, with = spicer_<flavor>()) and mash(brew, with = masher_<flavor>()). Hitting the tab key with your curser inside the parentheses of masher_flavor()will create a drop-down menu that shows a list of the arguments that go along with your brew's flavor.

If you have no trouble remembering the parameters that go along with your brew's flavor, or if you just want your code to be more concise, you don't have to use the with argument. Instead, you can just specify parameter values directly using the ... argument in the mash and spice functions. In the examples below, both approaches are shown.

masher_soft(
  bs = TRUE,
  bs_maxit = 20,
  bs_thresh = 1e-09,
  bs_row.center = FALSE,
  bs_col.center = TRUE,
  bs_row.scale = FALSE,
  bs_col.scale = TRUE,
  si_type = "als",
  si_thresh = 1e-05,
  si_maxit = 100,
  si_final.svd = TRUE
)

`bs`	a logical value. If `TRUE`, `softImpute::biScale()` is applied to `data_ref` or `rbind(data_ref, data_new)` prior to fitting `softImpute` models.
`bs_maxit`	an integer indicating the maximum number of iterations for the `biScale` algorithm.
`bs_thresh`	convergence threshold for the `biScale` algorithm.
`bs_row.center`	a logical value. If `TRUE`, row centering will be performed. If `FALSE` (default), then nothing is done.
`bs_col.center`	a logical value. If `TRUE` (default), column centering will be performed. If `FALSE`, then nothing is done.
`bs_row.scale`	a logical value. If `TRUE`, row scaling will be performed. If `FALSE` (default), then nothing is done.
`bs_col.scale`	a logical value. If `TRUE` (default), column scaling will be performed. If `FALSE`, then nothing is done.
`si_type`	two algorithms are implemented, type="svd" or the default type="als". The "svd" algorithm repeatedly computes the svd of the completed matrix, and soft thresholds its singular values. Each new soft-thresholded svd is used to re-impute the missing entries. For large matrices of class "Incomplete", the svd is achieved by an efficient form of alternating orthogonal ridge regression. The "als" algorithm uses this same alternating ridge regression, but updates the imputation at each step, leading to quite substantial speedups in some cases. The "als" approach does not currently have the same theoretical convergence guarantees as the "svd" approach.
`si_thresh`	convergence threshold for the `softImpute` algorithm, measured as the relative change in the Frobenius norm between two successive estimates.
`si_maxit`	maximum number of iterations for the `softImpute` algorithm.
`si_final.svd`	only applicable to `si_type = "als"`. The alternating ridge-regressions do not lead to exact zeros. With the default `final.svd = TRUE`, at the final iteration, a one step unregularized iteration is performed, followed by soft-thresholding of the singular values, leading to hard zeros.

a list with input values that can be passed directly into mash, e.g mash(brew, with = masher_nbrs()) for a neighbors brew or mash(brew, with = masher_soft()) for a soft brew.

x1 = rnorm(100)
x2 = rnorm(100) + x1
x3 = rnorm(100) + x1 + x2

outcome = 0.5 * (x1 - x2 + x3)

data <- data.frame(x1=x1, x2=x2, x3=x3, outcome=outcome)

n_miss = 10

data[1:n_miss,'x1'] = NA
sft_brew <- brew_soft(data, outcome=outcome, bind_miss = FALSE)

# these two calls are equivalent
mash(sft_brew, with = masher_soft(bs = FALSE))
mash(sft_brew, bs = FALSE)

knn_brew <- brew_nbrs(data, outcome=outcome, bind_miss = TRUE) %>%

# these two calls are equivalent
mash(knn_brew, with = masher_nbrs(fun_aggr_ctns = median))
mash(knn_brew, fun_aggr_ctns = median)