R Graphical Manual

Browse All

Last data update: 2014.03.03

R: The 'semdiag' function

semdiag

R Documentation

The `semdiag` function

Description

A function to calculate distance and parameter estimates for SEM diagnostics

Usage

semdiag(x,  EQSmodel, varphi=0.1, EQSdata='data.txt', D='E', delete=integer(0), max_it=1000, EQSprog='C:/Progra~1/EQS61/WINEQS',serial="111111 222222 333333", ram.path, software='EQS')

Arguments

`x`	Data frame or data matrix
`ram.path`	Ram path for R sem package
`software`	Software to be used. `sem` for R sem package and `EQS` for the use of EQS
`varphi`	Percentage of data to be down-weighted
`EQSmodel`	EQS input file name
`EQSdata`	Data file name used in EQS input file
`D`	How to treat prediction error. E: errors; F: factors
`delete`	A vector of data to be deleted. For example c(99,100) delete the 99th and 100th cases.
`max_it`	The maximum number of iterations
`EQSprog`	The path to where EQS program is installed.
`serial`	Serial no. for EQS. This is a string with spaces. Currently, it does not need to be supplied.

Value

`d_f`	Distance for f
`d_r`	Distance for r
`mu`	Mean
`p`	Number of observed variables
`q`	Number of factors used in calculating f and r
`res`	Model fit and paramter estimates. nml: normal ML; tsr: two-stagte robust; dr: direct robust
`eqs`	Full EQS output for the above three models
`x`	Data

Author(s)

Zhiyong Zhang and Ke-Hai Yuan

Maintainer: Zhiyong Zhang <zhiyongzhang@nd.edu>

References

Yuan, K.-H. and Zhang, Z. (2011). Structural Equation Modeling Diagnostics Using R Package semdiag and EQS. Structural Equation Modeling: An Interdisciplinary Journal.

Yuan, K.-H., and Hayashi, K. (2010). Fitting data to model: Structural equation modeling diagnosis using two scatter plots. Psychological Methods, 15, 335–351.

Examples

## Not run: 
## Examples based on EQS
library(semdiag)

## Example 1. Normally distributed data
data(N100)

## Not run
## The EQS input file is semplot.eqs
## Model 1: treating prediction errors as factors
N100out.1<-semdiag(N100, 'semplot.eqs', D='F')

## Diagnostics plot
semdiag.plot(N100out.1)

## Summary output 
semdiag.summary(N100out.1)

## Model 0: treating prediction errors the same as measurement errors
N100out.0<-semdiag(N100, 'semplot.eqs')

## Diagnostics plot
semdiag.plot(N100out.0)

## Summary output 
semdiag.summary(N100out.0)

## Example 2. Contaminated data
data(N85)

## The EQS input file is semplot.eqs
## Model 1: treating prediction errors as factors
N85out.1<-semdiag(N85, 'semplot.eqs', D='F')

## Diagnostics plot
semdiag.plot(N85out.1)

## Summary output 
semdiag.summary(N85out.1)

## Model 0: treating prediction errors the same as measurement errors
N85out.0<-semdiag(N85, 'semplot.eqs', D='E')

## Diagnostics plot
semdiag.plot(N85out.0)

## Summary output 
semdiag.summary(N85out.0)

## Case profile plot
semdiag.cpp(N85out.0, cases=c(86, 90, 98:100))

## Delete the 99th and 100th observations
N85out.1.del<-semdiag(N85, 'semplot.eqs', D='F', delete=c(99,100))

## Examples based on the sem package
library(sem)
data(N100)

## path diagram for the model
sem1<-specify.model()
f1 -> y1, NA, 1
f1 -> y2, l1, NA
f1 -> y3, l2, NA
f2 -> y4, NA, 1
f2 -> y5, l3, NA
f2 -> y6, l4, NA
f3 -> y7, NA, 1
f3 -> y8, l5, NA
f3 -> y9, l6, NA
f1 -> f2, g1, NA
f1 -> f3, g2, NA
f2 -> f3, g3, NA
y1 <-> y1, e1, NA
y2 <-> y2, e2, NA
y3 <-> y3, e3, NA
y4 <-> y4, e4, NA
y5 <-> y5, e5, NA
y6 <-> y6, e6, NA
y7 <-> y7, e7, NA
y8 <-> y8, e8, NA
y9 <-> y9, e9, NA
f1 <-> f1, e10, NA
f2 <-> f2, e11, NA
f3 <-> f3, e12, NA

## Model 1: treating prediction errors as factors
N100out.1<-semdiag(N100, ram.path=sem1, software='sem', D='F')

## Diagnostics plot
semdiag.plot(N100out.1)

## Summary output 
semdiag.summary(N100out.1)

## Model 0: treating prediction errors the same as measurement errors
N100out.0<-semdiag(N100, ram.path=sem1, software='sem')

## Diagnostics plot
semdiag.plot(N100out.0)

## Summary output 
semdiag.summary(N100out.0)

## Example 2. Contaminated data
data(N85)

## Model 1: treating prediction errors as factors
N85out.1<-semdiag(N85, ram.path=sem1, software='sem', D='F')

## Diagnostics plot
semdiag.plot(N85out.1)

## Summary output 
semdiag.summary(N85out.1)

## Model 0: treating prediction errors the same as measurement errors
N85out.0<-semdiag(N85, ram.path=sem1, software='sem', D='E')

## Diagnostics plot
semdiag.plot(N85out.0)

## Summary output 
semdiag.summary(N85out.0)

## Case profile plot
semdiag.cpp(N85out.0, cases=c(86, 90, 98:100))

## Delete the 99th and 100th observations
N85out.1.del<-semdiag(N85, ram.path=sem1, software='sem', D='F', delete=c(99,100))

## End(Not run)

The semdiag function