I want to output a path diagram of distributed covariance structure analysis (SEM) by linking Python and R.

Motivation

--I can't find a module to do sem with python! --I want to automatically output a beautiful sem diagram --I want to use R on python ――It is R (I feel) that the itchy part of the analysis can be reached. -According to R vs Python: Compare Data Analysis --Python is simpler for tasks other than statistics --R generally supports more statistics

** Theory: The strongest theory if supplemented with python and R **

--R and python can be linked by using "pyper" which is a python module.

Pyper settings on python

--Installing pyper pip install pyper

--Import pyper and create pyper indent --It is necessary to specify the path of R with "RCMD =" (with //) -By setting ʻuse_numpy ='True', use_pandas ='True'`, it is possible to read numpy, pandas type data into R.

python

import pyper 
import os
import numpy as np
import pandas as pd
#Create pyper indent
r = pyper.R(RCMD="C:\\Program Files\\R\\R-3.2.5\\bin\\x64\\R", use_numpy='True', use_pandas='True')

Creating sem program files in R

--At first, I was using the R "sem" package, but after a series of errors, I switched to "lavaan" (crying). --The sem function in lavaan was easier to set up. --For creating this file, refer to Try "First covariance structure analysis Amos path analysis" with lavaan. We received. --R has pre-installed packages "lavaan" and "semplot" to be used ――It seems that using this "semplot" will easily output a diagram of the estimation result of "lavaan". --This R program creates an SEM model, estimates it, and saves the path diagram as .png.

R(lavaan.R)

library(lavaan)
library(semPlot)

r <-  cor(dat)

model1 <- ('
  reach=~a1*Intermediate_test+a2*term_end_exam+a3*minitest
  skill=~b1*mock_examination1+b2*mock_examination2+b3*mock_examination3 
  reach~~skill
  ')

fit <- sem(model1, sample.cov=r, sample.nobs=284)


png("C:\\Users\\xxxxxxxx\\Documents\\Python Scripts\\python_R\\semplot.png ") 
semPaths(fit, "std", edge.label.cex = 1,line = 3, curvePivot = TRUE,
         sizeMan = 8, sizeInt = 1, sizeLat = 8)
dev.off()

Link python and R

--Let python read the data --This time, Analysis using R (SEM) of I converted "book usage data" into English and used it.

python

test = pd.read_csv("C:\Users\xxxxxxx\Documents\R\input\data\ch14sem.csv")

--Pass python data to R --Pass "test" to R as "dat".

python

r.assign("dat", test)

--Execute the R file "lavaan.R" with python

python

r("source(file='C:\\Users\\xxxxx\\Documents\\R\\input\\script\\lavaan.R')")

--r.get ("")can extract objects in R

python(Extraction of correlation matrix)

 r.get("r")

--It is possible to execute the R command on python on python with r ("") and output the result.

python(Estimated result display)

print r("summary(fit, standardized=TRUE, fit.measure=TRUE)")

Output path diagram

According to semPlot Examples | --Sacha Epskamp, this path diagram can be set in detail. (With this, I don't have to draw a path diagram with PowerPoint or draw.io. .)

Where it gets stuck

--If you don't indent on R, you will get an error on python even if you don't get an error on R. --Plot on R was not brought to python with r.get ("") , but .png was output. --Use // to specify the path.

The site that I used as a reference

Analysis using R (SEM)

http://mizumot.com/handbook/?page_id=625

Try "first covariance structure analysis path analysis by Amos" with lavaan

https://sites.google.com/site/officeoga/r/hajimeteno

Works with Python and R

http://qiita.com/ynakayama/items/f84dc659f1337d71dd9e

PypeR

http://www.webarray.org/softwares/PypeR/

R vs Python: Compare data analysis

http://postd.cc/r-vs-python-head-to-head-data-analysis/

semPlot Examples | - Sacha Epskamp

http://sachaepskamp.com/semPlot/examples