fwdmsa (version 0.2)

fwdmsa-package: Robust Mokken Scale Analysis by Means of the Forward Search Algorithm for Outlier Detection

Description

The package conducts the Forward Search on test and questionnaire data, and shows forward plots for the detection of outliers.

Details

Package:

fwdmsa

Type:

Package

Version:

0.2

Date:

2011-07-26

License:

GPL Version 2 or later

The package includes the functions

fs.MSA

Computes the necessary input for forward plots

plot.fs.class

S3 method for forward plots

fs.MSA.n1

Computes $n1$

plot.fs.n1.class

S3 method for a plot showing graphically $n1$

and data set

acs

Autonomy-Connectedness Scale

Thanks are due to L. Andries van der Ark for contributing R code, and Marrie Bekker and Marcel van Assen for providing the data set.

References

Bekker M. H. J., and Van Assen, M. A. L. M. (2006). A short form of the autonomy scale: Properties of the autonomy-connectedness scale (ACS-30). Journal of Personality Assessment, 86, 51-60.

Van der Ark, L. A. (2007). Mokken scale analysis in R. Journal of Statistical Software. http://www.jstatsoft.org

Zijlstra, W. P., Van der Ark, L. A., and Sijtsma, K. (2011). Robust Mokken scale analysis by means of the forward search algorithm for outlier detection. Multivariate Behavioral Research, 46, 58-89.

Examples

## Not run: 
# ## Analyses of Zijlstra et al. (2010).
# ## First Forward Search Analysis
#    library(fwdmsa)
#    data(acs.cont)
# 
#  # Determining n1 = 292
#  # Takes approximately 40 minutes
#    fs1.1.n1 <- fs.MSA.n1(acs.cont, B=100)
#    n1 <- fs1.1.n1$n1 
# 
#  # Figure 2: Plot of number unique subsamples
#    plot(fs1.1.n1)
# 
#  # Running the forward search
#    fs1.1 <- fs.MSA(acs.cont)
# 
#  # Figure 3: Plot of objective function
#    plot(fs1.1, type="objective", observations=1:618, col="gray70", n0=TRUE, n1=fs.res.cont.n1$n1, xlim=c(0,650))
#    plot(fs1.1, type="objective", id.observation=619, col=1, lwd=2, lty=2, add=TRUE)
#    plot(fs1.1, type="objective", observations=589:618, lwd=2, add=TRUE)
# 
#  # Figure 4: Gap plot
#    plot(fs1.1, type="gap", ylim=c(-10,12), n0=TRUE, n1=292)
# 
#  # Figure 5: Follow-up plots
#    plot(fs1.1, type="followup", step=543:548, reference.step=543, n0=TRUE, n1=292)
# 
# ## Remove influential observations from the data
#    acs.sus <- acs.cont[-(589:618),]
#  
#  # Determining n1 = 296
#    fs1.2.n1 <- fs.MSA.n1(acs.sus, B=100)
#    n1 <- fs1.2.n1$n1 
# 
#  # Running the forward search
#    fs1.2 <- fs.MSA(acs.sus)
# 
#  # Figure 6: Minexcl plot
#    plot(fs1.2, type="minexcl", n0=TRUE, n1=296, n2=TRUE)
# 
#  # Figure 7: Plot of number of scales
#    plot(fs1.2, type="num.scale", n0=TRUE, n1=296, n2=TRUE)
# 
#  # Figure 8: Item entry plot for the longest scale
#    plot(fs1.2, type="scale", id.scale=1, n0=TRUE, n1=296, n2=TRUE)
# 
# ## Second Forward Search Analysis
#  # Remove bad items from the data
#    acs.min.core <- acs.cont[-(589:618),-c(3,7,8,11,13,16)]
#  
#  # Determining n1 = 302
#    fs2.n1 <- fs.MSA.n1(acs.min.core, B=100)
#    n1 <- fs2.1.n1$n1 
# 
#  # Running the forward search
#    fs2 <- fs.MSA(acs.min.core)
# 
#  # Figure 9: Plot of restscore regression of item 1 for steps 302 and 589
#    plot(fs2, type="restscore", step=302, items=1, lty=2, ylim=c(0,4), n0=TRUE, n1=302, n2=TRUE)
#    plot(fs2, type="restscore", step=589, items=1, lty=1, add=TRUE)
# 
#  # Figure 10: Plot of estimated IRF of item 1
#    plot(fs2, type="IRF", items=1, n0=TRUE, n1=302, n2=TRUE)
# 
#  # Figure 11: Plot of coefH
#    plot(fs2, type="coefH", n0=TRUE, n1=302, n2=TRUE, ylim=c(.1,.8))
# 
# ## What if influential observations were not removed from the data
#    acs.cont.core <- acs.cont[,-c(3,7,8,11,13,16)]
#  # Determining n1 = 347
#    fs3.n1 <- fs.MSA.n1(acs.cont.core, B=100)
#    n1 <- fs3.n1$n1 
# 
#  # Running the forward search
#    fs3 <- fs.MSA(acs.cont.core)
# 
#  # Figure 12a: Plot of estimated IRF of item 1 with influential observations
#    plot(fs3, type="IRF", items=1, n0=TRUE, n1=347, n2=FALSE)
# 
#  # Figure 12b: Plot of coefH with influential observations
#    plot(fs3, type="coefH", n0=TRUE, n1=347, n2=FALSE, ylim=c(.1,.8))
# ## End(Not run)