Minor / low-priority. Inspired by reviewing CRAN's stabs, which accepts a formula + data interface in addition to x/y.
What cssr has today
css() (and the wrappers) take X as a numeric matrix or a data.frame (coerced via model.matrix) plus a separate y. There is no formula interface.
Proposed scope
An optional convenience formula interface, e.g. css(y ~ ., data = df, ...), dispatching to the existing matrix path. Purely ergonomic — no change to the method.
Notes
- Lowest priority of the stabs-inspired ideas; file for completeness.
- Would need to define sensible behavior for
clusters when columns come from a formula (cluster indices currently refer to columns of X).
- Consider whether an S3
css.formula / css.default split is worth the added surface, or whether documentation of the existing data.frame path is enough.
References
stabs: ?stabsel (formula vs matrix interface).- cssr:
css, checkCssInputs (X coercion).
Minor / low-priority. Inspired by reviewing CRAN's
stabs, which accepts aformula+datainterface in addition tox/y.What cssr has today
css()(and the wrappers) takeXas a numeric matrix or a data.frame (coerced viamodel.matrix) plus a separatey. There is noformulainterface.Proposed scope
An optional convenience
formulainterface, e.g.css(y ~ ., data = df, ...), dispatching to the existing matrix path. Purely ergonomic — no change to the method.Notes
clusterswhen columns come from a formula (cluster indices currently refer to columns ofX).css.formula/css.defaultsplit is worth the added surface, or whether documentation of the existing data.frame path is enough.References
stabs:?stabsel(formula vs matrix interface).css,checkCssInputs(X coercion).