The proof: your own published numbers, reproduced
I re-run your published analysis from public data (OSF / PLOS / GitHub) and hand over a side-by-side comparison. You see your numbers confirmed, or a discrepancy found — which may itself warrant a reproducibility report.
| Method family | Worked example | Result |
|---|---|---|
| Factorial ANOVA | Metadiscourse corpus study (PLOS ONE) | 9/9 F-values exact to 2 dp |
| CFA + measurement invariance | Quality-of-life scale study (Sci Rep 2023) | 13/15 exact + 2 pattern |
| Mediation / moderation | Emoji leadership study (PLOS ONE) | 5/5 PROCESS estimates exact |
| + 16 more audits | GLMM · network science · ML classifiers · latent growth · eye-tracking LMMs · VLM… | per-paper comparison tables |
Nineteen replication audits in total — each with a per-paper comparison table you can check in 15–30 minutes, across psychology, linguistics, education, and business studies.
These public-data replication audits are an audit before you hire: they show that I can recover published results, trace discrepancies, and document a reproducible workflow before you share your data. The US$1,000 (≈ HK$8,000) trial applies the same process to one bounded question from your paper and reproduces your numbers — a low-risk way to assess the work before commissioning a larger package.
What I can build into your grant
Working demos — script, data, report — adaptable to your data in days.
| If your research involves… | The pipeline I run | Result / status |
|---|---|---|
| EEG / ERP | Filtering → epochs → components | Prototype on single-subject tutorial data |
| Speech, phonetics | ASR + acoustic measures | Prototype on synthetic sample |
| Child speech | Child-Cantonese ASR + error analysis | SSD errors in carrier frames — not yet on real child recordings |
| Eye-tracking | Fixation/reading-time measures | Prototype on simulated data; LMMs scoped as a build deliverable |
| Interviews | LLM-assisted coding + reliability | κ 0.81 / 93% agreement |
| Reviews | Systematic-review screening | κ 0.88 vs manual |
| Surveys / scales | Reliability, factors, invariance | α .81–.85, 100% recovery |
| Experiments (RT) | Outliers → stats → figures | d = 0.50, matched |
| Corpus / text | NLP, sentiment, classifiers | 97% accuracy / κ 0.93 |
| Business text | Sentiment, scales, PROCESS | 5/5 PROCESS exact |
| Humanities | Corpus & discourse analysis | Discourse frame 61% vs 65% |
| Translation | Quality + terminology check | 100% preference, 90% error detection |
Every demo carries a data-status badge — real data, or prototype (tutorial / simulated / synthetic). See every demo →
No data yet? The pipeline starts before the data.
Literature & synthesis
Structured matrix from your reading list — design, sample, measures, findings — plus screened synthesis.
Evidence: 11 papers → one matrix, κ 0.76–0.88. Demo →
Instruments & experiments
Forward/back-translation checked against published Chinese; experiment materials and pre-registration.
Evidence: GAD-7/PHQ-9 items checked against the published Chinese; injected error caught. Demo →
Grant proposals
Draft scored against RGC criteria — where the panel will attack, and the patch.
Evidence: GRF2 2026/27 criteria, criterion-by-criterion. Demo →
See it. Try it. Build it.
1 · See it
A 15–30 minute walkthrough call for scholarly and technical discussion of methods, fit, and limitations. Every demo →
2 · Try it
Trial on your published data, checked against your paper. How it works →
3 · Build it
Full pipeline on current data — milestone-based, documented handover. Process →
Need admin, teaching or a private-assistant setup? FAQ →
How we work
Five steps: know the data → sanity-check → build in an open stack → verify (against the paper or the data) → report and hand over. Full account →
Contact
Email your paper or data link — no call needed. Indicative quote →
Pricing: One trial deliverable on your topic — from US$1,000 (≈ HK$8,000), free if not useful.
Replies within 24h — same day for urgent.
Working arrangement
Fixed-fee pricing — no hourly meter.