O’Connell & McCoach (Eds.) (2008): Multilevel Modeling of Educational Data

Today, Michael and I enjoy reading “Multilevel Modeling of Educational Data” edited by Ann A. O’Connell & D. Betsy McCoach. For anyone interested to becoming familiar with multi-level modeling, we especially recommend the following articles:

• “Modeling Growth Using Multilevel and Alternative Approaches” by Janet K. Holt (pp. 111-159): Janet gently explains data requirements and assumptions as well as forms of data (multivarite/wide data layout vs. person-period/long data layout). In addition, Table 4.5 (pp. 154-155) provides a consice overview to strengths and limitations of three alternative growth curve modeling methods.
• “Reporting Results from Multilevel Analyses” by John M. Ferron et al. (pp. 391-426): John and several other colleagues provides a valuable guideline and many suggestions for what to present when reporting results from multilevel analysis and how to present these results (text vs. tables vs. figures).
• “Software Options for Multilevel Models” by J. Kyle Roberts and Patrick McLeod (pp. 427-467): Kyle and Patrick provide a well organized, brief explanations of the most popular multilevel software packages (MLwiN, HLM, SAS, S-PLUS, R, SPSS, Mplus, and STATA).

London, UK: One-day Mplus training

A one-day training course titled “An introduction to Structural Equation Modelling using Mplus” to be held at City University, London, UK in December 2010

on Monday 13th December, 2010
and repeated on Tuesday 14th December, 2010

The price for the course is £250, with a student rate of £175

Details of the course are given below: for further information and to book a place, go to http://www.figureitout.org.uk

This course is also available on an inhouse basis; see http://www.figureitout.org.uk for more details.

Read more

New Book: Datenanalyse mit Mplus

Unfortunately, this posting concerns only German readers:

Christian Geiser’s book “Datenanalyse mit Mplus” is now published. We recommend this book to everyone using Mplus in Germany because this book contains not only a plethora of screenshots but also Mplus outfile tables. Further, the book is accompanied by several data files. Hip Hip Hooray to Christian!

Check out the Publishers Website for more details.

Bayes finally reached Mplus

In 2007, “Structural Equation Modelling: A Bayesian Approach” was published by Sik-Yum Lee. This monograph gave the first full account on Bayesian SEM. However, interested readers may struggle with the WinBUGS software introduced in the monograph. The good news is now: Four new documents describing Bayesian analysis in Mplus are now available! I especially recommend this article: Muthén, B. (2010). Bayesian analysis in Mplus: A brief introduction. Technical Report. Version 3.

Invariance testing using lavaan

Today, Michael showed me the lavaan function <measurement.invariance>. I instantly was delirious with joy: While in AMOS one struggles through several chi-square difference test tables, lavaan provides an easy to read output – no more, no less.

Measurement invariance tests:

Model 1: configural invariance
chisq      df  pvalue    cfi    tli  rmsea       bic
18.570 26.000   0.854  1.000  1.010  0.000  9979.522

Model 2: weak invariance (equal loadings):
chisq      df  pvalue    cfi    tli  rmsea       bic
25.096 31.000   0.763  1.000  1.007  0.000  9956.090

[Model 1 versus model 2]
delta.chisq  df  p.value  delta.cfi
6.53   5  0.25838     0.0000

Model 3: strong invariance (equal loadings + equal intercepts):
chisq       df  pvalue    cfi    tli  rmsea        bic
28.078  36.000   0.824  1.000  1.008  0.000  10012.996

[Model 1 versus model 3]
delta.chisq  df  p.value  delta.cfi
9.51  10  0.48469     0.0000

[Model 2 versus model 3]
delta.chisq  df  p.value  delta.cfi
2.98   5  0.70275     0.0000

Model 4: equal loadings + intercepts + means:
chisq      df  pvalue     cfi    tli  rmsea        bic
31.045  38.000    0.781  1.000  1.008  0.000  10003.979

[Model 1 versus model 4]
delta.chisq  df  p.value  delta.cfi
12.47 12  0.40837     0.0000

[Model 3 versus model 4]
delta.chisq  df  p.value  delta.cfi
2.97   2  0.22687     0.0000