Skip to contents

projak vs spmR

Technical Comparison Notes

2026-08-11

1 Purpose

  • projak (Ben Williams, v0.0.0.9000): pure-R package for NPFMC projection scenarios (1-7) downstream of an RTMB assessment; accepts a report object as input.
  • spmR (Jim Ianelli, v0.3.0): Standard Projection Model wrapper around the legacy ADMB spm executable with an experimental RTMB path; reads .dat inputs (ADMB-style).

2 Architecture

Component projak spmR
Input RTMB report object spm.dat plus species files
Compiled code None (pure R) ADMB spm.tpl plus partial RTMB stub
Multi-species support Single-species Multi-species natively (nspp)
Scenario coverage 1-7 fully implemented Alternatives 1-5 (RTMB path is a stub)
Maturity Early development (0.0.0.9000) More complete (tests, vignettes, pkgdown)

3 Recruitment

Both projects implement inverse-Gaussian recruitment, but with different parameterization paths:

  • projak: takes arithmetic and harmonic means directly from report$recruits.
  • spmR RTMB path: derives CV from the harmonic-to-arithmetic ratio, then samples recruitment.

4 Tier and Harvest Control Rule Logic

  • projak includes a full get_tier_f() implementation with Amendment 56 ramp logic, SSL protection, and scenario-specific F rules.
  • spmR RTMB path (runSPM_rtmb) currently stubs the scenario loop; SSB, catch, and F are placeholder means rather than forward age-structured projections.

5 Key Gap

runSPM_rtmb() in spmR is currently a scaffold: recruitment draws are implemented, but age-structured population dynamics are not. In contrast, projak::project_step() performs full age-structured projection (survival, catch-at-age, plus-group dynamics, and SSB updates).

6 Dependencies

  • projak: data.table, tidytable, magrittr
  • spmR: dplyr, ggplot2, ggthemes, patchwork, readr, stringr, tibble, tidyr

7 Summary

projak is effectively a focused, RTMB-native replacement for the projection component, with complete harvest control rule logic and lighter dependencies. spmR remains broader infrastructure (data I/O, multi-species SPM workflow, ADMB wrapper, and visualization), but its RTMB path still needs substantial development to match projak projection fidelity.