科研 Skill

神经连接性分析核查

neuro-connectivity-claims-check

Guides choice of leakage-robust connectivity measures such as wPLI, imaginary coherency and AEC, with confound controls. Use when reviewing or planning a connectivity analysis.

发布者 Codar安全审核通过v0.1.0更新于 2026-11-18

SKILL.md

Connectivity: choose robust measures and rule out artifacts

Two sensors see the same sources (volume conduction and field spread), share a reference, and differ in signal-to-noise. Each can create "connectivity" with no interaction at all. Choose measures and controls that remove these before interpreting.

1. Pick a measure that tolerates zero-lag mixing

QuestionUseAvoid on raw sensor data
phase couplingimaginary coherency (Nolte et al. 2004), wPLI (Vinck et al. 2011), PLI (Stam et al. 2007)coherence, PLV (inflated by zero-lag mixing)
amplitude couplingorthogonalised amplitude-envelope correlation (Hipp et al. 2012; Brookes et al. 2012)plain envelope correlation
directed influenceGranger/PDC/DTF on source data, with stationarity checks (Barnett & Seth 2011)Granger after aggressive filtering

Source-space analysis reduces but does not remove leakage; apply leakage correction there too.

2. Controls that every comparison needs

  • Reference: common reference creates shared signal; use a reference-free approach (bipolar for ECoG/sEEG, source space, or surface Laplacian for EEG) or show results hold across references.
  • Trial count: PLV, coherence and wPLI are biased by number of trials; equalise counts across conditions or use unbiased estimators (for example debiased wPLI).
  • Signal-to-noise and power: a condition with more power shows more phase consistency. Report power alongside connectivity and test whether the effect survives when power is matched.
  • Surrogates: compare with time-shifted or phase-randomised surrogates to get a null per edge.
  • Multiple comparisons: all channel pairs x frequencies; use network- based statistics or permutation with max statistic.

3. Interpretation limits

Connectivity is statistical dependence, not anatomical connection; a common driver produces coupling between areas that do not talk. Graph metrics depend strongly on thresholding and network density; compare networks at matched density.

Output

Measure and why, reference and leakage handling, trial and power matching, null model, correction method, and the claim scoped to what the measure supports.

示例提示词

帮我检查这份EEG相干性和wPLI连接分析的方法选择是否合理。

所属插件

神经科学面向神经科学研究:EEG/MEG、fMRI、BIDS、spike sorting、连接组、脑机接口解码评估、神经刺激安全与公开神经数据检索。

安装该插件会一并获得同领域的 45 个 Skill / MCP。

同插件的其他科研 Skill

常见问题

神经连接性分析核查能做什么?

Guides choice of leakage-robust connectivity measures such as wPLI, imaginary coherency and AEC, with confound controls. Use when reviewing or planning a connectivity analysis。

如何安装神经连接性分析核查?

在科搭 Codar 左侧点击「插件」,在「组件」页签搜索「神经连接性分析核查」并安装;在 Claude Code 或 Codex 中可以发送本页的安装提示词,或下载 Skill 包后解压到 ~/.claude/skills 或 ~/.codex/skills。

神经连接性分析核查需要哪些权限?

该科研 Skill没有声明额外权限;执行具体操作时仍由 Agent 按自身规则请求确认。

神经连接性分析核查经过官方审核吗?

是。SpeedSkills 收录的每个版本都会记录内容摘要与权限声明,当前版本已通过安全审核。