Skip to contents

Getting help with acir

Source: SUPPORT.md

Reading first

Three articles ship with the package, and between them they answer most questions.

  • Assimilative causal inference (ACI) in R walks the whole workflow on a worked example.
  • The closed-form ACI engine documents the machinery underneath: how models are written, what the filter, smoother and online-smoother arguments do, and which conventions each reported number is computed under.
  • Reproducing the reference MATLAB codebase records which quantities are graded against the authors’ implementation, where the package departs from it deliberately, and how far each correspondence has been checked.

vignette(package = "acir") lists them. Function documentation is available through ?aci, ?aci_range and the rest.

Asking a question

Open an issue at https://github.com/biometryhub/ACI/issues.

A question about a result is easiest to answer with the model, the parameters and the observed output. A short reproducible example is ideal, and aci_simulate() takes a seed, so a simulated case can be made reproducible in one line.

Reporting a numerical disagreement

These are the most valuable reports this package receives, and they are treated as findings rather than as support requests. The issue tracker offers a template for them.

Please include the system, the parameters, the quantity that disagrees, and the size of the disagreement. If you have compared against another implementation, say which one and at what settings. Conventions differ between implementations in ways that look like disagreements and are not, so naming the settings matters.

tools/oracle/parity/ in the repository holds the harness used to grade this package against the authors’ reference, and inst/evidence/register.csv records the tolerance each graded quantity meets, so a disagreement can be read against the tolerance it is supposed to sit inside.

Reporting a defect

Also the issue tracker. A defect report is most useful with the output of sessionInfo(), the call that produced the problem, and what you expected instead.