# Methods & Limitations — Enhanced Studies Cycle-Log Aggregates (v1.0.0)

## What this is
An anonymized, aggregate summary of **487 self-reported "cycle logs"** across
**50 performance-enhancement compounds**, harvested from public bodybuilding /
harm-reduction forums, then reduced to **derived statistics only**. It is a
real-world-use signal, **not a clinical study**.

## How it was built
1. **Harvest** — public forum threads discussing personal use were collected per compound.
2. **Per-log extraction** — dose, duration, stack, gender, and a plain-language outcome
   summary were extracted (deterministic regex for dose/duration/gender; see `derive_v3.py`).
3. **Aggregation** (`aggregate_logs.py`) — per compound: gender split; dose distribution
   (median/min/max of the reported figure, grouped by unit); duration distribution
   (normalized to weeks); co-reported compounds (alias-matched); source-domain provenance.
   Every sparse field carries an explicit **coverage %**.
4. **Outcome classification** — each log's free-text outcome was classified into
   worked / mixed / stopped-or-regretted, and explicitly-reported side effects tallied,
   by a language model constrained to the log text (batched per compound).
5. **Anonymization gate** — the published artifacts contain **derived counts only**:
   no usernames, no verbatim post text, no raw free-text. An automated leak-check
   asserts no handle or quote fragment survives into the output.

## Limitations (read before citing)
- **Self-report** from anonymous forum users; no verification, no clinical oversight,
  no adverse-event surveillance. Side-effect counts reflect what users *mentioned*,
  not incidence.
- **Small n** (6–12 logs/compound). Treat every figure as indicative, not representative.
- **Selection bias** — people who post logs (and troubleshoot problems) are not a random
  sample. Example: Testosterone logs skew toward TRT bloodwork-management threads, so its
  outcomes read "mixed" despite Testosterone's established efficacy.
- **Dose is not frequency-normalized** — a reported "250 mg" is the figure stated, not a
  weekly total (logs inconsistently record EOD/weekly schedules).
- **Outcome/side classification** is model-assisted and inherits the text's ambiguity;
  "mixed" is used when unclear.

## Ethics & use
Aggregated, anonymized, no personal data. Educational/informational only; **not medical
advice** and not an endorsement of use. Each compound's live evidence page:
https://enhancedstudies.com/compound/<slug>/. License: CC-BY-4.0. Never fabricated:
where evidence is thin, the numbers say so.