如何在Python中用REML实现嵌套式Variant Components Analysis(VCA)

我尝试用以下Python代码通过REML进行变异分量分析（Variant Components Analysis, VCA）：

import pandas as pd
import statsmodels.api as sm
from statsmodels.formula.api import ols
data = {'Part':[1,1,1,1,1,1,1,1,1,2,2,2,2,2,2,2,2,2,3,3,3,3,3,3,3,3,3],
        'Employee':[1,1,1,2,2,2,3,3,3,1,1,1,2,2,2,3,3,3,1,1,1,2,2,2,3,3,3],
        'Measurement':[103.3, 103.1, 103.1, 103.3, 102.9, 103.6, 103.2, 103.6, 103.1, 104.5, 104.8, 103.9, 104.5, 104, 103.8, 103.8, 103.6, 104, 104, 103.6, 103.5, 103.9, 104.1, 104.5, 104.3, 103.9, 103.8]} 
df = pd.DataFrame(data)
lm = ols('Measurement ~ C(Part) + C(Part):(Employee)', data=df).fit()
table = sm.stats.anova_lm(lm, typ=2)
table['Percentage of Total Variance'] = (table['sum_sq'] / table['sum_sq'].sum()) * 100

但通过JMP的指定步骤（粘贴表格、选择Analyze> Quality and Process > Variability / Attribute Gauge Chart，设置REML分析等）以及R的VCA包代码：

library('VCA')
library('dplyr')
df <- read.csv('simpler.csv', sep=';')
df$Part <- as.factor(df$Part)
df$Employee <- as.factor(df$Employee)
fitREML <- remlVCA(form=Measurement~Part+Part/Employee, Data = df)

得到了一致的嵌套变异分量结果，Python代码的输出却无法匹配。请问是否可以不使用rpy2在Python中实现符合要求的分析？

需求细节 #

• 需采用REML分析
• 需进行嵌套分析
• 曾尝试statsmodels混合模型，但分组语法不清晰

内容的提问来源于stack exchange，提问作者Ranieri Bubans