Assemble user-facing prediction deliverables.

Builds two CSVs in .path.export: predictions.csv (wide best per response, optional Transformed twin) and audit.csv (long disaggregated per method, raw values).

Usage

postProcess(
  .path.summary,
  .path.export,
  key,
  responses,
  location,
  locationSources = c("data/domains/index.csv", file.path(.path.export, "X.csv"),
    file.path(.path.export, "Xo.csv")),
  transform = NULL,
  normalize = NULL,
  boundsMin = -Inf,
  boundsMax = Inf
)

Arguments

.path.summary: Directory holding ssel computed aggregates (response_long.csv, residuals_oof.csv, metrics.csv, prediction_quantiles.csv).
.path.export: Directory holding ssel intermediate Yo.csv and where outputs (predictions.csv, audit.csv) are written. Created if missing.
key: Identity column name (typically "SampleID").
responses: Character vector of response column names.
location: Character vector of location columns to carry from the project's identity source into predictions.csv / audit.csv.
locationSources: Character vector of candidate file paths from which to read (key, location). The first file carrying all required columns is used; otherwise the union of all matching files is taken.
transform: Element-wise transform spec, a list with op and op-specific params. Supported: list(op="identity") (default when NULL), list(op="log", base=10), list(op="exp", base=exp(1)).
normalize: Row-wise normalize spec. Supported: list(op="identity") (default), list(op="closure", target=100).
boundsMin: Lower clamp for predicted values. Default -Inf.
boundsMax: Upper clamp for predicted values. Default Inf.

Value

Invisible NULL. Side effect: writes predictions.csv and audit.csv to .path.export.