power-analysis-guide

本条目面向分析数据与指标归因场景,提供“power-analysis-guide”技能的目录入口。公开标签:learning、样本量、效应量、统计功效。请通过固定上游版本核对实际功能、依赖与权限;此页为 Skillπ 独立编写的用途介绍和原站链接,不代表运行安全保证。原文与附件依各自许可开放。

仓库
brycewang-stanford/Auto-Empirical-Research-Skills
仓库最近推送
Stars
5,023
Forks
578
Issues / PRs
0

安装方式

先用 Prompt 核对来源,也可切换到安装命令或下载可用的安装包。

安装方式

用 AI 助手安装

复制提示词,让助手先核对固定版本和依赖。

查看 Prompt收起 Prompt
先检查当前 Agent 是否已安装 SkillPI Skill Finder。如果尚未安装,请按官方仓库的说明安装并启用:https://github.com/Findskills/skillpi。然后使用 Skill Finder 打开此 SkillPI 详情页并定位 "power-analysis-guide":https://skillpi.com/zh/skills/deepskill-93ab92bd549ba3f8226d355980977f960e14b063fbce421f8798b64113648d11。安装前核对上游来源和版本:https://github.com/brycewang-stanford/Auto-Empirical-Research-Skills/blob/bfa0a6e5a843babbf45a1ef1663fd5dc4f07c0ba/skills/43-wentorai-research-plugins/skills/analysis/statistics/power-analysis-guide/SKILL.md。来源不一致时停止并说明差异。核对通过后,用 Skill Finder 安装目标 Skill,并报告安装路径和来源版本。

此命令从 GitHub 来源仓库获取指定技能。运行前请先查看源文件。

运行工具
npx skills add https://github.com/brycewang-stanford/Auto-Empirical-Research-Skills/tree/bfa0a6e5a843babbf45a1ef1663fd5dc4f07c0ba/skills/43-wentorai-research-plugins/skills/analysis/statistics/power-analysis-guide --skill power-analysis-guide
许可协议: —

来源与配置

查看 GitHub 源码

brycewang-stanford/Auto-Empirical-Research-Skills固定提交 bfa0a6e5a843babbf45a1ef1663fd5dc4f07c0ba

配置方式待核实

请查看原始来源中的适用工具、依赖和权限要求。

安装实测待补充

本条目面向分析数据与指标归因场景,提供“power-analysis-guide”技能的目录入口。公开标签:learning、样本量、效应量、统计功效。请通过固定上游版本核对实际功能、依赖与权限;此页为 Skillπ 独立编写的用途介绍和原站链接,不代表运行安全保证。原文与附件依各自许可开放。

↓ 0 下载 ↗ 0 分享点击

文件浏览

在原站浏览文件 ↗

原始 SKILL.md

打开原文 ↗
---
name: power-analysis-guide
description: "Sample size calculation and statistical power analysis guide"
metadata:
  openclaw:
    emoji: "🎯"
    category: "analysis"
    subcategory: "statistics"
    keywords: ["sample size calculation", "power analysis", "effect size", "significance testing"]
    source: "wentor-research-plugins"
---

# Power Analysis Guide

Calculate appropriate sample sizes for your study using power analysis, understand effect sizes, and avoid underpowered or wastefully overpowered designs.

## Core Concepts

### The Four Parameters of Power Analysis

Every power analysis involves four interrelated quantities. Fix any three to solve for the fourth:

| Parameter | Symbol | Definition | Typical Value |
|-----------|--------|-----------|---------------|
| **Effect size** | d, r, f, etc. | Magnitude of the phenomenon you expect to detect | Varies by field |
| **Significance level** (alpha) | alpha | Probability of Type I error (false positive) | 0.05 |
| **Statistical power** (1 - beta) | 1 - beta | Probability of detecting a true effect | 0.80 or 0.90 |
| **Sample size** | N | Number of observations needed | Solve for this |

### Error Types

| | H0 is true (no effect) | H0 is false (effect exists) |
|---|---|---|
| **Reject H0** | Type I error (alpha) | Correct (power = 1 - beta) |
| **Fail to reject H0** | Correct (1 - alpha) | Type II error (beta) |

## Effect Size Conventions

### Cohen's d (Two-Group Comparison)

```
d = (M1 - M2) / SD_pooled
```

| Size | Cohen's d | Interpretation |
|------|-----------|---------------|
| Small | 0.2 | Subtle, may need large N to detect |
| Medium | 0.5 | Noticeable, typical in social sciences |
| Large | 0.8 | Obvious, often visible without statistics |

### Correlation (r)

| Size | r | r-squared |
|------|---|-----------|
| Small | 0.1 | 1% variance explained |
| Medium | 0.3 | 9% variance explained |
| Large | 0.5 | 25% variance explained |

### Cohen's f (ANOVA)

| Size | f | Equivalent eta-squared |
|------|---|----------------------|
| Small | 0.10 | 0.01 |
| Medium | 0.25 | 0.06 |
| Large | 0.40 | 0.14 |

### Odds Ratio (Logistic Regression)

| Size | OR |
|------|-----|
| Small | 1.5 |
| Medium | 2.5 |
| Large | 4.0 |

## Power Analysis in Python (statsmodels)

### Two-Sample t-Test

```python
from statsmodels.stats.power import TTestIndPower

analysis = TTestIndPower()

# Solve for sample size
n = analysis.solve_power(
    effect_size=0.5,    # Cohen's d = medium
    alpha=0.05,         # Significance level
    power=0.80,         # 80% power
    ratio=1.0,          # Equal group sizes
    alternative='two-sided'
)
print(f"Required N per group: {int(n) + 1}")  # Output: 64

# Solve for power (given N)
power = analysis.solve_power(
    effect_size=0.5,
    alpha=0.05,
    nobs1=50,
    ratio=1.0,
    alternative='two-sided'
)
print(f"Power with N=50 per group: {power:.3f}")  # Output: 0.697
```

### Paired t-Test

```python
from statsmodels.stats.power import TTestPower

analysis = TTestPower()
n = analysis.solve_power(
    effect_size=0.3,    # Small-medium effect
    alpha=0.05,
    power=0.80,
    alternative='two-sided'
)
print(f"Required N (paired): {int(n) + 1}")  # Output: 90
```

### One-Way ANOVA

```python
from statsmodels.stats.power import FTestAnovaPower

analysis = FTestAnovaPower()
n = analysis.solve_power(
    effect_size=0.25,   # Cohen's f = medium
    alpha=0.05,
    power=0.80,
    k_groups=4          # Number of groups
)
print(f"Required N per group: {int(n) + 1}")  # Output: 45
```

### Chi-Square Test

```python
from statsmodels.stats.power import GofChisquarePower

analysis = GofChisquarePower()
n = analysis.solve_power(
    effect_size=0.3,    # Cohen's w = medium
    alpha=0.05,
    power=0.80,
    n_bins=4            # Degrees of freedom + 1
)
print(f"Required total N: {int(n) + 1}")
```

### Multiple Regression

```python
from statsmodels.stats.power import FTestPower

analysis = FTestPower()
# For R-squared: convert to f2 = R2 / (1 - R2)
r_squared = 0.10  # Expected R-squared for the model
f2 = r_squared / (1 - r_squared)  # f2 = 0.111

n = analysis.solve_power(
    effect_size=f2,
    alpha=0.05,
    power=0.80,
    df_num=5            # Number of predictors
)
# n returned is df_denom; total N = n + df_num + 1
total_n = int(n) + 5 + 1
print(f"Required total N: {total_n}")
```

## Power Analysis in R (pwr Package)

```r
library(pwr)

# Two-sample t-test
result <- pwr.t.test(d = 0.5, sig.level = 0.05, power = 0.80,
                     type = "two.sample", alternative = "two.sided")
cat("N per group:", ceiling(result$n), "\n")

# Correlation test
result <- pwr.r.test(r = 0.3, sig.level = 0.05, power = 0.80,
                     alternative = "two.sided")
cat("Total N:", ceiling(result$n), "\n")

# One-way ANOVA (4 groups)
result <- pwr.anova.test(k = 4, f = 0.25, sig.level = 0.05, power = 0.80)
cat("N per group:", ceiling(result$n), "\n")

# Chi-square test
result <- pwr.chisq.test(w = 0.3, df = 3, sig.level = 0.05, power = 0.80)
cat("Total N:", ceiling(result$N), "\n")

# Plot power curve
result <- pwr.t.test(d = 0.5, sig.level = 0.05, power = NULL,
                     n = seq(10, 200, by = 5))
plot(result)
```

## Using G*Power (Desktop Application)

G*Power (gpower.hhu.de) is a free, widely-used GUI application for power analysis:

1. **Select test family**: t-tests, F-tests, chi-square, z-tests, exact tests
2. **Select statistical test**: e.g., "Means: Difference between two independent means (two groups)"
3. **Select type of analysis**: A priori (compute N), Post hoc (compute power), Sensitivity (compute detectable effect)
4. **Input parameters**: Effect size, alpha, power, allocation ratio
5. **Calculate**: Click "Calculate" to get the result
6. **Plot**: Use "X-Y plot for a range of values" to visualize power curves

## Practical Recommendations

### Choosing Effect Sizes

Do NOT blindly use Cohen's conventions. Instead:

1. **Literature review**: Find effect sizes reported in similar studies
2. **Pilot data**: Run a small pilot study to estimate the effect
3. **Smallest effect of interest (SESOI)**: What is the smallest effect that would be practically meaningful?
4. **Meta-analyses**: Use pooled effect sizes from meta-analyses in your area

### Common Mistakes

| Mistake | Problem | Solution |
|---------|---------|----------|
| Post hoc power analysis | Circular and uninformative after data collection | Only do a priori power analysis |
| Using Cohen's "medium" by default | May be unrealistic for your field | Base on literature or SESOI |
| Ignoring attrition | Actual N may be lower than planned | Inflate N by 10-20% for expected dropout |
| Forgetting multiple comparisons | Bonferroni corrections reduce power | Adjust alpha for the number of tests |
| Not reporting power analysis | Reviewers cannot evaluate adequacy | Always report in Methods section |

### Reporting Template

```
A priori power analysis was conducted using [G*Power 3.1 / statsmodels / R pwr].
For a [test name] with an expected effect size of [d/r/f = X] (based on
[source: previous study / meta-analysis / pilot data]), alpha = .05, and
power = .80, the required sample size was [N per group / total N]. To account
for an estimated [X]% attrition rate, we recruited [final N] participants.
```

版本历史

  1. v2026.07 GitHub ↗
    原版更新说明

    ## Highlights - **First tagged release.** Every gate is green: `make check` (validate + 183 unit tests + eval-harness + numeric benchmarks), validate-catalog, quality-evals, and OpenSSF Scorecard on `main`. - **Rigor double loop at 13 method families.** Each family closes taxonomy tag → eval scenario (does the agent avoid the method's known traps?) → deterministic numeric benchmark (does it recover the true effect?). New this cycle: heterogeneous treatment effects (CATE) and quantile treatment effects (QTE). - **Paper-WorkFlow flagship at 29/29 executable rigor gates**, now merged to its `main` with a drift-gated README badge — the credibility artifact is checked by CI, so it can never silently go stale ([RIGOR.md](https://github.com/brycewang-stanford/Paper-WorkFlow/blob/main/RIGOR.md)). - **1,150 vendored skills / 69 collections**, all license-audited, security-scanned, tracked in a machine-readable catalog with CI-enforced freshness, and importable as a single whole-repo router skill. - **Hardened weekly upstream syncs**: synced trees must pass the full gate before merge; this cycle both sync PRs were verified, patched (StatsPAI's SkillOpt execution-gate card restored), and merged. - **First-party plugin marketplace** (Python / R / Stata empirical-analysis plugins + guide) shipped in the previous cycle, now under a tagged version. ## Known Follow-Ups - Ecosystem benchmark comparison (AERS vs econometrics agents on the shared benchmark suite) — [PLAN-2026-07 week 3](https://github.com/brycewang-stanford/Auto-Empirical-Research-Skills/blob/main/docs/PLAN-2026-07.md). - Decomposition (Oaxaca–Blinder) and shift-share (Bartik) method families (weeks 3–4). - GitHub Pages catalog site, README demo screenshots, five-language README consistency script (week 4). ## Catalog - Top-level collections: **69** - `SKILL.md` files: **1150** - Cataloged agent tools: **335** ([`tools/CATALOG.md`](../tools/CATALOG.md)) ## Methodological rigor - Method families in the coverage map: **13** ([`RIGOR_COVERAGE.md`](RIGOR_COVERAGE.md)) - Eval scenarios: **26** (138 rubric items, 122 auto-checkable) ([`../eval-harness/`](../eval-harness/README.md)) - Benchmark tasks: **13** (70 gold checks, 58 required) ([`../benchmark/`](../benchmark/README.md)) ### Method families (skills tagged) - `iv`: 27 - `rdd`: 25 - `panel-fe`: 19 - `staggered-did`: 17 - `bayesian`: 13 - `did`: 10 - `event-study`: 10 - `matching`: 10 - `synthetic-control`: 10 - `cate`: 8 - `dml`: 8 - `survival`: 3 - `quantile`: 2 ### Eval scenarios by category - `causal-identification`: 14 - `citation-hygiene`: 1 - `reproducibility`: 2 - `research-integrity`: 1 - `runtime-safety`: 1 - `statistical-validity`: 2 - `writing-compliance`: 3 - `writing-style`: 2 ## License buckets - AGPL-3.0: 1 - Apache-2.0: 1 - CC-BY-NC-4.0 (non-commercial): 1 - CC-BY-SA-4.0: 1 - CC-BY-SA-4.0 (repository default): 4 - GPL-3.0: 1 - MIT: 35 - MIT Non-Commercial: 1 - Mixed (per-skill; see folder): 2 - UNKNOWN - check upstream: 22 ## Quality gates - `make check` — catalog/provenance/audit/eval freshness, repo hygiene, Paper-WorkFlow gates, stdlib unit tests, eval-harness lint + graded smoke, and the strict numeric benchmark (reference-gold drift fails the build). - External links and OpenSSF Scorecard run on their scheduled workflows.

当前收录版本

0.0.1

核对来源版本 ↗

质量评测

满分 100。前四项来自 DeepSkill 收录的质量报告。

场景价值92/100
评分依据

解决科研和数据分析中的核心痛点:样本量计算与统计功效分析

触发精准85/100
评分依据

触发词覆盖全面:sample size calculation, power analysis, effect size, significance testing

结构规范95/100
评分依据

结构完整:包含核心概念、参数说明、代码示例、实用建议

执行可靠90/100
评分依据

工具来源明确:statsmodels、R pwr包、G*Power等主流工具

证据可验 (前四项均值)91/100
计算方式

取前四项原始分值的算术平均,四舍五入取整。可通过收录报告核对各项评分与说明,运行记录需单独查证。

权限与使用前检查

根据当前主文档自动整理,命中内容可能是示例或防护说明;不是实际权限声明,也未覆盖附件、外部依赖或运行行为。

未从主文档识别到上述常见操作,不能据此判断无需权限或没有风险。

本站展示提示不会授予权限。涉及文件改动、密钥、外部发送或费用时,仍由你使用的工具按需确认。

安全评级与访问范围

risk · 静态证据核验,不代表运行安全保证。

使用前请由宿主核对并授权实际文件读写、外部服务和联网目的地;本站的核验不会授予任何运行权限。

来源与使用声明 SkillPi 的部分技能资料来自公开来源或创作者提交。第三方内容的权利归各自权利人所有。安装或运行前,请查看原站的许可、权限要求和使用风险。

来源与使用

请按详情页的作者、原始来源和许可证核对使用条件,运行前检查技能需要的权限。目录收录不等于已验证安装或安全性。