CODEWITHSUNDEEP
pythonrstatsmodelssas-jmp
Ranieri Bubans
Ranieri Bubans

Reputation: 129

How to perform a Variant Components Analysis (VCA) in Python using Reduced Maximum Likelihood? (analogous to R VCA and JMP Variance Components)

I have the following Python code where I've already try to perform a VCA analysis using REML:

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

And then the following result is generated: Python anova result

When I run this Análises in JMP by running the steps:

  1. Paste the table
  2. Select Analyze> Quality and Process > Variability / Attribute Gauge Chart
  3. Set Part and Employee as 'X, grouping' variables and Measurement as 'Y, response'
  4. Set Analysis Settings options to 'Use REML analysis'
  5. Run the analysis and generate the Nested Variance Components The result is the following: jmp variance components

And the same result could be achieved in R by the following code:

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)

Rresult

However I cannot get the same result in Python. Is there a way that it could be done in Python without using rpy2?

Detail 1: REML analysis is required for the data that I'm working on.
Detail 2: Nested analysis is required according to R and Python formulas.
Detail 3: I've already tried using mixed models from statsmodels, but the group syntax is not clear.

Upvotes: 6

Views: 463

Answers (0)

Related Questions