```{mermaid}
%%| label: fig-flowchart-pipeline
%%| fig-cap: "A flowchart of the full research process. All steps described are further explained in the methodologic report and the supplements."
flowchart TD
    A["Population
    Top 100 JIF journals 2013-2023"]

    B["Crossref 
Metadata filtering
95,042 → 40,860 publications
Deduplication, date filter, keyword exclusions"]

    C["Precision-based stratified sampling
Target ±1.5 pp · n ≈ 4,265"]

    D["Sample A
n = 408 · unstratified
SI classifier training only"]

    E["Sample B
analytical sample
n = 4,265 · stratified by year"]

    F["Manual + LLM labelling
Subset of Sample A
κ ≈ .83 after reconciliation"]

    G["SI classifier trained
Random Forest / XGBoost
TF-IDF keyword features"]

    H1["Full-text retrieval
HTML / PDF, scraped"]
 
    H2["Full-text retrieval
HTML / PDF, scraped"]

    I["SI classifier applied to Sample B"]

    I2["SI papers identification
n = 1,763 with usable full text"]

    OA["OA classified from metadata
using Crossref, Web of Science, Scopus"]

    J["OSP training subset
n = 352, from SI papers in Sample B
manual & LLM labelling"]

    KOD["OD classifier
RF / XGBoost"]

    KOM["OM classifier
RF / XGBoost"]

    KPR["Preregistration classifier
RF / XGBoost"]

    L["Prevalence estimates, Post-stratified by year, adjusted for misclassification"]

A --> B
B -- Training/Testing Sample --> D
B -- Analytical Sample --> C
D --> H1 --> F --> G
C --> E --> H2 --> I
G --> I
E -- OA from metadata --> OA
I --> I2 --> J
J --> KOD & KOM & KPR
KOD & KOM & KPR -- Applied to all SI papers --> L
OA --> L
J --> L
```


![A flowchart of the full research process. All steps described are further explained in the methodologic report and the supplements found in the OSF repository.](img/research-flow.svg){#fig-flowchart-pipeline width=100%}