From c3f09b737a2bc2c139f3de5c8cbca874220e4185 Mon Sep 17 00:00:00 2001 From: =?utf8?q?J=C3=A9r=C3=B4me=20Benoit?= Date: Fri, 3 Oct 2025 23:55:08 +0200 Subject: [PATCH] chore: refine github copilot instructions MIME-Version: 1.0 Content-Type: text/plain; charset=utf8 Content-Transfer-Encoding: 8bit Signed-off-by: Jérôme Benoit --- .github/copilot-instructions.md | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/.github/copilot-instructions.md b/.github/copilot-instructions.md index 00f655d..72a5623 100644 --- a/.github/copilot-instructions.md +++ b/.github/copilot-instructions.md @@ -24,11 +24,13 @@ These instructions guide GitHub Copilot to generate changes consistent with this ## Statistical conventions +- **Hypothesis testing**: use a single test statistic (e.g., t-test) when possible. - **Divergence metrics**: document direction explicitly (e.g., KL(A||B) vs KL(B||A)); normalize distributions; add epsilon to avoid numerical issues. - **Distance vs divergence**: distinguish clearly; use consistent terminology. - **Effect sizes**: report alongside test statistics and p-values; use standard formulas; document directional interpretation. - **Distribution comparisons**: use multiple complementary metrics (parametric and non-parametric). -- **Correlations**: prefer robust methods; report confidence intervals when feasible; document assumptions. +- **Correlations**: prefer robust estimators; report the correlation estimate with a confidence interval (parametric or bootstrap) when feasible. +- **Uncertainty quantification**: use confidence intervals or credible intervals methods when feasible. - **Normality tests**: combine visual diagnostics (e.g., QQ plots) with formal tests when assumptions matter. - **Multiple testing**: document corrections or acknowledge their absence. -- 2.43.0