Research Methodology

Hypothesis-driven structure

Articles begin with a defined question and identify the evidence or worked example used to discuss it. A hypothesis-and-test structure may explain a method, but QTJ does not present an illustrative calculation as a preregistered empirical study.

Data

Articles name published sources when they rely on external facts or records. References to a vendor, market series, or database describe the source framework; they do not imply that QTJ owns or downloaded a proprietary dataset unless a reproducible artifact is linked.

Statistical inference

Worked examples may use bootstrap confidence intervals, multiple-testing corrections, or Bayesian intervals to explain how a claim should be evaluated. Parameters and outputs remain illustrative unless the article links the underlying data and calculation.

Backtest methodology

When discussing a backtest, QTJ explains the intended train/test split, refit cadence, rebalance frequency, transaction-cost assumptions, and lookahead-bias controls. This is a review standard, not a claim that every article contains an executed QTJ backtest.

Transaction costs

Where transaction costs would affect an interpretation, the article should state the spread, commission, slippage, or market-impact assumption. Worked-example cost figures are sensitivity inputs, not live execution quotes.

Robustness checks

Articles discuss the robustness checks a practitioner should seek: alternate windows, parameter sensitivity, shifted samples, different universes, and explicit failure conditions. Where QTJ has not run the check on linked data, the discussion is methodological rather than empirical.

Reproducibility

Formulae, parameters, and data windows are included when they help a reader reproduce a worked example. A result is described as reproducible only when the source data and calculation path are available to the reader.