Learning how to fit latent growth curve in Amos (created 2010-08-25).

This page is moving to a new website.

I've run a few structural equation models (SEM) using AMOS software in the past, but it's been a while. I might have to fit some more AMOS models in the future, so I took some time to reacquaint myself with the program. I thought I'd try something a bit different than the tutorial examples and something a bit new for me, a latent growth curve model. It didn't work, but I thought I'd document what I tried here. Then when I figure out what is going on, possibly with the help of some colleagues who are smarter than I am, I'll update this page.

The first thing to do was to try to find some data. My two favorite sites for data are the Data and Story Library (DASL) and the Australian counterpart (OzDASL).

• Matthew Hutcheson, Mike Meyer, Cara Olson, Paul Velleman, John Walker, Cornell University. The Data and Story Library (DASL). Excerpt: "DASL (pronounced 'dazzle') is an online library of datafiles and stories that illustrate the use of basic statistics methods. We hope to provide data from a wide variety of topics so that statistics teachers can find real-world examples that will be interesting to their students." This website was last verified on 2008-URL: lib.stat.cmu.edu/DASL/
• Gordon Smyth. Australasian Data and Story Library (OzDASL). Description: This website offers a library of data sets and associated stories. It is intended as a resource for teachers of statistics, and emphasis is given to data sets with an Australasian context. This website was last verified on 2008-URL: www.statsci.org/data/

There was an interesting data on growth patterns of wallabies at OzDASL, but this didn't seem like a good example to start with since the observation times were different for each animal. I then started hunting for the famous growth curve data set that I remembered from graduate school. I didn't find it, but I found something close, data on ramus mandibular growth in rhesus monkeys, which is shown in a paper by Schneiderman et al (1992), available at:

• http://deepblue.lib.umich.edu/bitstream/2027.42/30933/1/0000603.pdf

Here's reproduction of the raw data

I have picked up a book for my Kindle on Latent Growth Curve Modeling,

• Dr. Nancy E. Briggs. Latent Growth Curve Modeling. 1st ed. Sage Publications, Inc; 2008. Excerpt: "Latent growth curve modeling (LGM)-a special case of confirmatory factor analysis designed to model change over time-is an indispensable and increasingly ubiquitous approach for modeling longitudinal data. This volume introduces LGM techniques to researchers, provides easy-to-follow, didactic examples of several common growth modeling approaches, and highlights recent advancements regarding the treatment of missing data, parameter estimation, and model fit. The book covers the basic linear LGM, and builds from there to describe more complex functional forms (e.g., polynomial latent curves), multivariate latent growth curves used to model simultaneous change in multiple variables, the inclusion of time-varying covariates, predictors of aspects of change, cohort-sequential designs, and multiple-group models. The authors also highlight approaches to dealing with missing data, different estimation methods, and incorporate discussion of model evaluation and comparison within the context of LGM. The models demonstrate how they may be applied to longitudinal data derived from the NICHD Study of Early Child Care and Youth Development (SECCYD)."

for ten bucks and started reading it in five minutes. It had a diagram of a simple model, which I am reproducing here.

That seemed simple enough to do in Amos, except for that triangle. So here's what I built in Amos.

It didn't work and my feeble attempts to tweak the model failed. The missing triangle is the problem, I suspect. In the graphic world of SEM, a triangle represents a constant. There are no triangles in Amos and I don't know how to fit a constant properly.

I suspect I am overlooking something obvious, and I'll read a bit and ask around.

Here's an explanation of how to use AMOS for  latent growth curve models:

• http://ssc.utexas.edu/software/faqs/amos#Amos_4

and an image that shows a diagram similar to what I am working on.

I'll see if I can adapt my AMOS model to fit this format.