Maximal Returns Protocol (MRP)

Analysis code for a time-efficient concurrent training intervention — 12 weeks, 1 hour/week, combining HIIT + resistance training + HIFT plyometrics. Part of Dr. Jacob Bowie's dissertation work at the University of Connecticut.

Weekly progression across the 12-week intervention. (a) Per-exercise load progression with significance markers across six RT lifts. (b) Total weekly repetitions. (c) HIIT intensity (% Wmax). Color: trained (red) vs untrained (blue); shape: female (circle) vs male (triangle).

Study design

• Intervention: 12 weeks, one 60-minute session per week. Each session = 15 min HIIT (cycle, 5 × 1 min at 85% Wmax) + 15 min single-set-to-failure resistance training across six multi-joint exercises + 10 min HIFT plyometric circuit.
• Participants: n = 9 recreationally active adults (4 aerobically trained, 5 untrained) tested at PRE, MID, and POST.
• Funding: NASA Connecticut Space Grant Consortium (P-1708).

Key findings

• Significant within-subject strength gains across all 12 strength measures (PRE → POST p < 0.001).
• Effect sizes ranged from moderate to very large (Hedges' g = 0.49–1.54).
• Lower-body lifts showed the largest adaptations (10RM Deadlift g = 1.54).
• VO₂max improved in aerobically trained participants (p = 0.016, g = 0.68).
• Maximal power output (Wmax) increased progressively across the intervention.
• Running economy and body composition were maintained.

Given n = 9 with no control group, inferential results are treated as exploratory and hypothesis-generating rather than confirmatory. The analysis is framed to inform powering of a future controlled trial.

Repo layout

MRP/                              # repo root = project root
├── 01_MRP_Analysis.Rmd           # Canonical analysis (Phases 0–10 refactor)
├── MRP.Rproj                     # RStudio project anchor
├── MRP.xlsx                      # Anonymized analytic dataset
│                                 #   (sheets: Strength, vo2data,
│                                 #   Exercise Session RT, HIIT, HIFT)
├── renv.lock                     # Dependency lock file
├── helpers/
│   ├── paths.R                   # Absolute-path anchoring on MRP.Rproj
│   ├── setup.R                   # Packages, seed, namespace conflicts
│   └── functions.R               # fmt_p, fmt_g, create_styled_kable,
│                                 #   Hedges' g, MRP color scales, labels
├── renv/                         # renv infrastructure (library/ gitignored)
├── legacy/                       # Earlier exploratory scripts (see legacy/README.md)
├── DATA/                         # Raw subject-level data — NOT TRACKED
├── LICENSE                       # MIT (code)
├── CITATION.cff                  # GitHub citation widget
└── README.md

Rendering the analysis

# Open MRP.Rproj in RStudio (working directory = repo root)
renv::restore()                         # restore package versions
rmarkdown::render("01_MRP_Analysis.Rmd")

The rendered 01_MRP_Analysis.html is self-contained. YAML parameters (show_all_plots, save_pdfs, save_outputs, cache) can be overridden via render(params = ...) for batch runs that emit polished xlsx tables alongside the HTML.

Analysis structure

The canonical Rmd is organized around five research questions, each mapping to a labeled section:

1. Within-subject change (PRE → POST) — mixed-effects models per outcome (lmer with subject random intercept), Tukey-adjusted contrasts via emmeans.
2. Magnitude of adaptation — Hedges' g with small-sample correction, plus % change per measure.
3. Trained vs untrained response — Timepoint × TrainingStatus interaction; flagged as exploratory secondary given subgroup sizes.
4. Intermediate trajectory — PRE / MID / POST shape for measures with mid-intervention testing.
5. Dose–response within RT — cumulative weekly volume-load progression vs individual strength gain.

Two sensitivity analyses follow the primary results:

• Sex-adjusted refit of every primary LMM (adds Sex as a fixed effect), comparing Type III p-values against the unadjusted models and flagging α = .05 crossings.
• Baseline-adjusted ANCOVA (POST ~ TrainingStatus + PRE) across all PRE/POST outcomes, with a parallel-slopes diagnostic and β(PRE) reported per measure to make regression-to-the-mean visible.

Design notes

• Tables use cell_spec conditional formatting via fmt_p() / fmt_g(): significant p-values in blue, moderate-or-larger effect sizes in red. Never column_spec(..., bold = TRUE) (silently overrides per-cell styling).
• Figures use a custom blue palette (#00205B / #6699CC / #A0C9E0) for PRE / MID / POST.
• Hand-written prose paragraphs are preserved alongside auto-generated, inline-computed prose blocks (blue-bordered divs) — useful for diffing the two truth sources during reviewer revision.
• Caching is disabled project-wide (cache: FALSE) — repeatedly bit by stale cache masking real bugs during refactor.
• Project root is anchored on MRP.Rproj via helpers/paths.R; chunk working directories are absolute, so knitr's per-chunk WD restoration cannot break path resolution.

Provenance of earlier scripts

The legacy/ directory contains dated .Rmd and .R files from earlier stages of the analysis (script.R, Session_script.R, chatGPT_script.R, TvU Script.R, 03142025 MRP_Markdown.Rmd, etc.). These are preserved as a historical record of how the analysis evolved, not as parts of the canonical pipeline. They contain hard-coded local paths (C:/MRP, G:/My Drive/...) from the original Windows + Google Drive workflow and are not expected to run on a clean clone — use 01_MRP_Analysis.Rmd for any actual rendering. See legacy/README.md for a file-by-file index.

Data

• MRP.xlsx (tracked) is the IRB-approved anonymized analytic dataset. Subjects are coded SUB <n>; no direct identifiers.
• DATA/ (not tracked, gitignored) holds raw per-subject material — consent forms, physician clearance, PARQ, HR/HRV exports, Polar JSONs. These remain on the lab share for provenance.
• HR/HRV / Polar-session data from two participant cohorts has been recovered into DATA/ for a future scope expansion; it is not used by the analyses in this repo.

Reproducibility

• R 4.5.2 (last verified 2026-04-09).
• Package versions pinned in renv.lock; restore with renv::restore().
• Reproducibility seed: set.seed(0308) in helpers/setup.R (applied before any stochastic call).

Status

The MRP manuscript was submitted to the Journal of Strength and Conditioning Research as a Research Note (Sept 2025). The journal returned a priority-fit rejection (Feb 2026) with constructive methods-level feedback — reframe as exploratory pilot / proof-of-concept, expand methods, drop overreach to military / spaceflight populations. The analyses in this repo (including the sex-adjusted + ANCOVA sensitivity additions) implement that revision plan; resubmission to an alternative outlet is in progress.

Cite

If you use this analysis pipeline or build on its design, please cite the repository via the GitHub "Cite this repository" widget (sourced from CITATION.cff) or:

Bowie, J. (2026). Maximal Returns Protocol — analysis pipeline. [Software]. https://github.com/JacobBowie/MRP

License

• Code — MIT (see LICENSE).
• MRP.xlsx — anonymized analytic data, available for academic non-commercial use; please cite the repository if used.
• Raw subject-level data in DATA/ is not distributed; contact the author for collaboration inquiries subject to IRB approval.