MethodVahti
← Vahtian

A product of Vahtian

Defend your sampling decision — honestly.

No-one arrives with tidy parameters. You arrive with pressure: a protocol to write, an N to defend, a reviewer’s comment, half-coded pilot interviews, or a nagging doubt about saturation. MethodVahti starts from your situation and helps you build a defensible, COREQ/SRQR-grounded argument. You’re not optimising a number — you’re defending a decision; the models run underneath.

Consensus Contested Author hypothesis Opinion range

Start from where you are Pick the situation you’re actually in — MethodVahti elicits the COREQ/SRQR detail it needs from there.

“I’m writing the protocol and need an N I can defend.”

You get
A planned N as a range from three models, a saturation expectation, and the methods paragraph for your protocol.
It asks
Your orientation, how varied the sample is, target analytic depth, and any pilot signal.
Our stance
A defensible range with the reasoning — not a single magic number.

What you walk away with A methods document you can submit — not a terminal artifact.

Whatever route you take, the output is the same shape: submission-ready methods text and a COREQ/SRQR-mapped record of every decision — for an ethics application, a protocol, a reviewer-response letter, or a manuscript. Every claim carries its epistemic status (◆ ◇ ○ ◌).

SectionStandards covered
Study profileSRQR 1–6 · COREQ 1–9
Sampling argument + optimisationSRQR 13 · COREQ 17
Three-model sample-size comparisonCOREQ 17
Fuzzy-set calibration sensitivitySRQR 5
Epistemic limitationsSRQR 20 · COREQ 31–32
COREQ 32-item checklistCOREQ complete
Citation + audit recordReproducibility

Download the sample report (PDF) → Try the live simulations →

The sample is illustrative — an IPA study of treatment decision-making in advanced NSCLC. Your report is built from your situation; the simulations let you feel how the models respond before you commit.

What runs underneath The computation stays under the argument — you never tune it by hand.

Primary ○

Hierarchical score

Heterogeneity estimated within each dimension, then aggregated across them. This is the number that feeds the sample-size argument.

Diagnostic ◆

Marginal map

A descriptive, per-dimension breakdown — which dimensions actually carry the variation, as a plain unweighted mean.

Diagnostic ○

Sparse-interaction stress

The full cross-tabulation with a sparsity penalty, so data gaps are visible — never silently driving the result.

Severity weight ≠ observed rate. An outcome weight of 0.90 means near-maximum amplification severity, not "90% disagreement". Defaults are author hypotheses; teams audit them with pilot data, and every change is timestamped in the report. Curious how the models move? Open the explorer →

Access Free for individuals during beta — not free forever, and not free for institutions.

free beta

Individual researchers

Free while in beta

  • Start from any of the five situations
  • The planning explorer + guided methods drafting
  • Beta users keep beta terms for the project they start under it

Individual pricing arrives at 1.0. Beta is the time to use it free and tell us what's wrong with it.

Start from your situation →
project licence

Institutions & teams

Paid — per research project

  • Departments, doctoral schools, research groups
  • Team-wide licence, unlimited iterations
  • Priority support for methods reviews
Talk to us →

Individuals start free from any situation and the planning explorer. The institutional project licence and guided drafting are how the work gets funded. Developers: the Apache-2.0 core is on GitHub.

Open validation, in progress

We don't hide that the primary heterogeneity score is an author hypothesis (○) — Vahtian's construction, not externally validated. We are testing it against open-access qualitative datasets and published meta-syntheses, and we will publish the validation protocol and results openly. Until that work lands, treat the score as decision support, not ground truth. Read the validation plan →

Honest limits: MethodVahti does not infer causality, does not validate study quality, and does not replace researcher judgment. The number is a synthesis you defend — not a verdict it issues.