Add unified design and transformation API (lhs_design / transform_lhs)

[bertcarnell]

Summary

Adds a high-level, "batteries-included" interface so users can go from a parameter specification to a ready-to-use data.frame in one call, instead of manually generating a [0,1] design and applying qnorm/qfactor/qinteger/qdirichlet column by column. This is a usability layer on top of the existing functionality — no new sampling math.

• lhs_design(n, variables, type = "random", ...) — generates a Latin hypercube sample of the requested type (random, improved, maximin, optimum, genetic, or oalhs) and transforms each margin, returning a named data.frame. The number of LHS columns is inferred from the variables spec, and ... is forwarded to the underlying generator (e.g. dup, maxSweeps).
• transform_lhs(lhs, variables) — applies the same margin transformations to an existing [0,1] design (e.g. from augmentLHS, or any matrix on [0,1]), so the transformation layer is reusable on its own.

Variable specification

Each named element of variables describes one variable and may be:

Spec	Example	Built on
a function of p	function(p) qexp(p, rate = 2)	—
a named quantile fn	list(dist = "qnorm", mean = 2, sd = 0.5)	match.fun + do.call
a factor	list(type = "factor", levels = c("a","b","c"), ordered = FALSE)	qfactor()
an integer range	list(type = "integer", min = 5L, max = 17L)	qinteger()
a Dirichlet block	list(type = "dirichlet", alpha = c(2,3,4))	qdirichlet() (consumes/produces length(alpha) columns)

set.seed(1234)
lhs_design(20, variables = list(
  x     = list(dist = "qnorm", mean = 2, sd = 0.5),
  grp   = list(type = "factor", levels = c("a", "b", "c")),
  count = list(type = "integer", min = 5L, max = 17L),
  z     = function(p) qexp(p, rate = 2),
  comp  = list(type = "dirichlet", alpha = c(2, 3, 4))
), type = "maximin")

What's included

• R/lhs_design.R — lhs_design, transform_lhs, and internal helpers, with roxygen docs
• man/lhs_design.Rd (documents both exported functions)
• NAMESPACE exports
• tests/testthat/test-lhs_design.R — covers every margin type, dispatch over all generator types, ... pass-through, validation, and reuse on an augmented design
• NEWS, README (.Rmd + .md) entries, updated Description, version bump to 1.4.0

Verification

R is not available in the environment used to prepare this branch, so the testthat suite was not executed here — please run R CMD check / devtools::test() before merging. The logic was reviewed carefully (column accounting, naming, and dispatch traced by hand). The man page was hand-written to match roxygen2 output; re-running roxygen2::roxygenise() should reproduce it. The README.md example shows only deterministic output (dimensions and column classes) since I could not knit it.

Note on coordination with the sliced/nested PR

This branch is independent of #71 and also bumps the version to 1.4.0 and edits NEWS/DESCRIPTION/README. If both are merged there will be small, easy-to-resolve conflicts in those metadata files (the R code does not overlap). Happy to rebase whichever merges second, or renumber the versions, however you prefer.

https://claude.ai/code/session_01JNVgdYk9zNB8B4hLSn8rzj

---

Generated by Claude Code