What EDGAR is

EDGAR stands for Experimental Design Generator and Randomiser. It was originally developed as a suite of Excel workbooks by the Biometrics team at Rothamsted Research. The same algorithms were re-implemented in the open-source Python project rotsl/edgar, distributed as edgar-design on PyPI. This R package is a native R port of that Python implementation. Python is not required at runtime.

Installing

Once the package is available on CRAN, install it with:

install.packages("ExperimentalDesignGeneratorandRandomiser")

For development, you can install from a local checkout with:

# devtools::install("/path/to/edgar-r")

Listing designs

library(ExperimentalDesignGeneratorandRandomiser)
list_designs()
#>       key                              name has_layout has_blocks
#> 1   cr_eq Completely Randomised, Equal...      FALSE      FALSE
#> 2 cr_uneq Completely Randomised, Uneq...     FALSE      FALSE
#> 3     rcb           Randomised Complete ...       TRUE       TRUE
#> ...

Generating a design

Use generate_design(type, ..., seed = 0L) with one of the nine design keys: cr_eq, cr_uneq, rcb, rcb_uneq, two_factor_rcb, latin, split_plot, variable_blocks, alpha.

res <- generate_design("rcb", treatment_count = 4, block_count = 3, seed = 42)
print(res)

Each design is also available via a design-specific convenience function:

res <- design_rcb(treatment_count = 4, block_count = 3, seed = 42)

Reproducibility

The package ports CPython’s Mersenne Twister seeding algorithm and Fisher-Yates shuffle to native R. The same integer seed produces the same design in R and in the upstream Python edgar-design package. Generating a design never modifies the global .Random.seed, so unrelated user code that uses sample() or runif() is not affected.

# Run twice with the same seed; the output is identical
res1 <- generate_design("rcb", treatment_count = 4, block_count = 3, seed = 42)
res2 <- generate_design("rcb", treatment_count = 4, block_count = 3, seed = 42)
identical(as.data.frame(res1), as.data.frame(res2))
#> [1] TRUE

# Different seeds produce different designs (with overwhelming probability)
res3 <- generate_design("rcb", treatment_count = 4, block_count = 3, seed = 43)
identical(as.data.frame(res1)$Variety, as.data.frame(res3)$Variety)
#> [1] FALSE

Working with the result

Every design returns an edgar_design S3 object. You can:

• print it,
• coerce it to an ordinary data.frame,
• inspect the metadata via $design_name, $parameters, $seed, $warnings, $generated_at,
• inspect the layout view via $layout, $layout_headers, $layout_section_labels, or via as_layout_frames().

df <- as.data.frame(res)
head(df)

Exporting

CSV export uses no extra dependencies:

write_edgar_csv(res, file = "design.csv")

JSON export requires the jsonlite package (in Suggests):

write_edgar_json(res, file = "design.json")

XLSX export requires the openxlsx package (in Suggests):

write_edgar_xlsx(res, file = "design.xlsx")

Provenance

EDGAR was originally developed by the Biometrics team at Rothamsted Research as Excel workbooks, available at edgarweb.org.uk. The algorithms were subsequently re-implemented in Python by the rotsl/edgar project, distributed as edgar-design on PyPI. This R package is a native R port of that Python implementation, with byte-identical cross-language reproducibility for the same integer seed. Alpha designs follow the methodology described by Patterson and Williams (1976).