imuRUN

Routine-Use Notebooks (imuRUN) is a spreadsheet-first front-end to the imuGAP model-fitting package. It is built for people who are comfortable running a command at the R prompt but are not power R users: describe the analysis in one Excel workbook, validate it with spreadsheet-referenced messages, fit with imuGAP::sampling(), and read the coverage estimates back in the same workbook.

Features

• Two input modes. Use a single .xlsx workbook or a directory containing observations, locations, and target files (CSV or RDS). imuRUN derives imuGAP's population rows from each observation's location, reference cohort, age span, and dose.
• Bundled template and example workbooks. imurun_init() writes a blank template with instructions and sampler configuration; imurun_copy_example() writes a complete example derived from imuGAP's *_sim data.
• Friendly validation. Inputs are checked against the canonical schema ([IMURUN_SCHEMA]) via a layer over imuGAP's canonicalizers that names the offending sheet, column, or row and collects every problem it can find rather than stopping at the first.
• Validate-only mode. run_fit(c("-h", <input>)) checks the workbook without fitting.
• Human-readable results. On success, the input workbook gains a results sheet containing medians and credible intervals beside the request context; fit.rds is also saved for advanced post-processing.
• Scriptable. The engine functions (run_fit(), read_inputs(), validate_inputs(), read_workbook(), and friends) are exported for use directly from R. The optional CLI wrapper returns shell exit codes (0 success, 1 validation, 2 model, 3 I/O) for use in pipelines.

Installation

imuRUN is a beta release; install it from GitHub for now:

# install.packages("remotes")
remotes::install_github("ACCIDDA/imuRUN")

Once it is on CRAN you will be able to install the released version with:

install.packages("imuRUN")

imuRUN depends on imuGAP; fitting is delegated to imuGAP::sampling(), which requires imuGAP's Stan-based model backend (a working Stan toolchain). Reading and writing .xlsx workbooks uses openxlsx2.

The R functions are the primary interface. To additionally make imuRUN available as a shell command, install the bundled wrapper onto your PATH:

imuRUN::install_cli()          # symlinks into ~/.local/bin (Unix) or writes
                               # an imurun.cmd shim (Windows)

Make sure the target directory (~/.local/bin by default) is on your PATH. If you would rather not install a launcher, you can always invoke the engine from R with imuRUN::run_fit(...).

Usage

run_fit() accepts a workbook path (or a directory of CSV/RDS inputs). The generated workbook's configuration sheet holds iter, chains, seed, and warmup; automation flags may override those values.

Walkthrough

1. Get a workbook to fill in. Scaffold a blank template into the current directory:

   imuRUN::imurun_init(".")

   This writes imurun_template.xlsx with instructions, configuration, observations, locations, and target sheets. To start from a filled, runnable example, use imuRUN::imurun_copy_example(".").

2. Fill it in. Enter sampled counts on observations, the hierarchy on locations, and prediction requests on target. Review the sampler values on configuration. The workbook instructions define every column.

3. Validate. Check the inputs without fitting:

   imuRUN::run_fit(c("-h", "imurun_template.xlsx"))

   Any problems are reported all at once, in spreadsheet terms. Validation does not require the Stan toolchain.

4. Fit. Once validation passes, run the model:

   imuRUN::run_fit("imurun_template.xlsx")

   imuRUN adds a results sheet to that workbook and writes fit.rds beside it. It refuses to replace existing results unless --overwrite is supplied.

The same steps work with a directory of CSV/RDS files in place of the workbook, for example run_fit("data") where data/ contains observations.csv, locations.csv, and target.csv.

From R

Every step is also available as an exported function, so you can drive the same pipeline from a script:

library(imuRUN)

# Copy the bundled example next to your work and inspect it
example <- imurun_copy_example(tempdir())

# Read and validate without fitting
inputs <- read_inputs(example)
validate_inputs(inputs)

# Or run the whole pipeline (amends the workbook and writes fit.rds)
run_fit(example)

See ?run_fit, ?read_inputs, ?validate_inputs, and ?IMURUN_SCHEMA for details, and the package website for the full reference.

Related

imuRUN is part of the imu* family and fronts imuGAP, the underlying model-fitting package.