PyStatPower¶

PyPI - Python Version GitHub License PyPI - Status

PyStatPower 是一个统计学功效分析的 Python 软件包，可用于样本量、检验效能和效应量大小的估计。

📚 功能模块¶

两独立样本均值非劣效检验
单样本率置信区间
单样本率差异性检验
两独立样本率差异性检验
两独立样本率非劣效检验
两独立样本率优效性检验
相关系数检验

📦 安装¶

前置需求：Python 3.10+

pip install pystatpower

🔨 使用¶

估算样本量¶

单样本率置信区间

from pystatpower.models import proportion

size = proportion.single.ci.solve_size(
    proportion=0.9,
    ci_width=0.10,
    alpha=0.05,
)
print(size)

# output: 158

单样本率差异性检验（单组目标值法）

from pystatpower.models import proportion

size = proportion.single.inequality.solve_size(
    null_proportion=0.80,
    proportion=0.95,
    alternative="one-sided",
    alpha=0.025,
    power=0.8,
)
print(size)

# output: 42

两独立样本率非劣效检验

from pystatpower.models import proportion

size = proportion.independent.noninferiority.solve_size(
    treatment_proportion=0.95,
    reference_proportion=0.90,
    margin=-0.10,
    ratio=1,
    alpha=0.025,
    power=0.8,
)
print(size)

# output: (48, 48)

计算检验效能¶

from pystatpower.models import proportion

power = proportion.independent.noninferiority.solve_power(
    treatment_proportion=0.95,
    reference_proportion=0.90,
    margin=-0.10,
    treatment_size=48,
    reference_size=48,
    alpha=0.025,
)
print(power)

# output: 0.800282915718918

反推效应量¶

from pystatpower.models import proportion

treatment_proportion = proportion.independent.noninferiority.solve_treatment_proportion(
    reference_proportion=0.90,
    margin=-0.10,
    treatment_size=48,
    reference_size=48,
    alpha=0.025,
    power=0.8,
)
print(treatment_proportion)

# output: 0.9499637015276098

🧪 兼容性测试结果¶

	🐍 3.10	🐍 3.11	🐍 3.12	🐍 3.13	🐍 3.14
SciPy 1.7	✅	-	-	-	-
SciPy 1.8	✅	-	-	-	-
SciPy 1.9	✅	-	-	-	-
SciPy 1.10	✅	✅	-	-	-
SciPy 1.11	✅	✅	✅	-	-
SciPy 1.12	✅	✅	✅	-	-
SciPy 1.13	✅	✅	✅	-	-
SciPy 1.14	✅	✅	✅	-	-
SciPy 1.15	✅	✅	✅	✅	-
SciPy 1.16	-	✅	✅	✅	✅
SciPy 1.17	-	✅	✅	✅	✅

注： - 表示该 Python 版本下不存在对应的 SciPy 发行版。

🔮 鸣谢¶