评估分布选择

Python 中的蒙特卡洛模拟

Izzy Weber

Curriculum Manager, DataCamp

选择变量的概率分布

直观理解数据与可用的概率分布
用最大似然估计（MLE）比较候选分布
用 Kolmogorov–Smirnov 检验评估分布的拟合优度
- 量化数据经验分布与候选理论分布的距离
- 用 scipy.stats.kstest() 计算

评估分布选择：年龄

results = []

list_of_dists = ["laplace", "norm", "expon"]

for i in list_of_dists:
    dist = getattr(st, i)

    param = dist.fit(dia["age"])

    result = st.kstest(dia["age"], i, args=param)

    print(result)

按拉普拉斯、正态、指数分布的顺序结果：

KstestResult(statistic=0.09511179937112832, pvalue=0.0006239579389182981)
KstestResult(statistic=0.0615913626181368, pvalue=0.06703225234359811)
KstestResult(statistic=0.2536037941921312, pvalue=1.5202547969084796e-25)

评估分布选择：年龄

按拉普拉斯、正态、指数分布的顺序结果：

KstestResult(statistic=0.09511179937112832, pvalue=0.0006239579389182981)
KstestResult(statistic=0.0615913626181368, pvalue=0.06703225234359811)
KstestResult(statistic=0.2536037941921312, pvalue=1.5202547969084796e-25)

评估分布选择：总胆固醇（血清）

results = []
list_of_dists = ["laplace", "norm", "expon"]

for i in list_of_dists:
    dist = getattr(st, i)
    param = dist.fit(dia["tc"])
    result = st.kstest(dia["tc"], i, args=param)
    print(result)

按拉普拉斯、正态、指数分布的顺序结果：

KstestResult(statistic=0.06435779928393615, pvalue=0.04915329841106708)
KstestResult(statistic=0.051165295747227724, pvalue=0.19085587687385897)
KstestResult(statistic=0.3318461436889846, pvalue=7.018486943525e-44)

Let's practice!

Python 中的蒙特卡洛模拟