Skip to content

README: regression-metamodel EVPPI and Gaussian-approximation EVSI (Jalal and Alarid-Escudero, 2018) - #259

Merged
feralaes merged 1 commit into
mainfrom
readme-voi
Nov 17, 2026
Merged

feralaes merged 1 commit into
mainfrom
readme-voi

Conversation

@feralaes

Copy link
Copy Markdown
Collaborator

Fernando asked for the README to promote dampack's value-of-information methods, in particular the Gaussian approximation approach to EVSI that he developed with Hawre Jalal:

Jalal H, Alarid-Escudero F. A Gaussian Approximation Approach for Value of Information Analysis. Medical Decision Making. 2018;38(2):174-188. doi:10.1177/0272989X17715627

Changes

  1. What it does: the VOI bullet now names the methods and credits them:
    • EVPPI from regression metamodels of the PSA (Strong et al., 2014; Jalal and Alarid-Escudero, 2018). These are the references ?calc_evppi already cites.
    • EVSI with the Gaussian approximation (Jalal and Alarid-Escudero, 2018).
    • Both work from the existing PSA sample, with no further model runs.
  2. New section, "Value of research: EVPPI and EVSI from one PSA":
    • why EVPPI and EVSI matter for research decisions;
    • the nested simulations they would normally need;
    • how the two dampack functions avoid them, describing the Gaussian approximation in two steps, regression metamodel then preposterior approximation, and its advantages from the paper (correlated parameters, non-Gaussian distributions, several parameters informed by one study).
  3. A runnable example on cancer_psa_data at $100,000/QALY, about 4 s to render. The sentence after it is computed inline:
    • EVPI: $22,684 per patient;
    • the EVPPI of the two treatment-failure probabilities is 36% of that ($8,213);
    • the EVSI of a new study rises from $1,331 at 10 patients to $7,895 at 1,000, approaching the EVPPI.
      A hidden check stops the README from rendering unless the EVSI rises with n and EVSI < EVPPI < EVPI.
  4. Citation: the Gaussian approximation paper is listed after the expected loss curves paper.

No new figure, so the installed size is unchanged.

Notes

Verification

  • devtools::build_readme() renders, and its checks pass.
  • R CMD check --as-cran: only the development-version NOTE, and the installed size is OK.

…tion EVSI

- 'What it does' credits the methods: EVPPI from regression metamodels (Strong et al. 2014;
  Jalal and Alarid-Escudero 2018) and EVSI with the Gaussian approximation approach (Jalal and
  Alarid-Escudero 2018, doi:10.1177/0272989X17715627), both from the existing PSA sample.
- New 'Value of research: EVPPI and EVSI from one PSA' section with a runnable example on
  cancer_psa_data (EVPI, EVPPI of two failure probabilities, EVSI by study size), with numbers
  computed inline and a hidden check that EVSI rises with n and EVSI < EVPPI < EVPI.
- The Gaussian approximation paper is listed under Citation.
@feralaes
feralaes merged commit a1f3d64 into main Nov 17, 2026
13 checks passed
@feralaes
feralaes deleted the readme-voi branch November 17, 2026 19:42
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant