Selon Adrian Musters <Adrian.Musters@NURS.UTAH.EDU>:

> I have been using the cluster analysis tools in SPSS and have recently > come across some software called Latent Gold 3.0.11 > (http://www.statisticalinnovations.com/products/latentgold.html). Is > there anyone who has used the software and can make recommendatations as > to whether or not it is better than the Cluster Analysis features in > SPSS?

Haven't looked at the clustering tool of SPSS recently, but latent class analysis (LCA) is something entirely different from hierachical and k-means clusters, even though it also groups cases.

But this is where the similarities end. It's more like factor analysis (FA) or structural equation models for non-interval-scaled variables. Actually, it is sort of a subtype of log linear models.

LCA is "superior" to FA, in that it allows for ordinally and nominally scaled data and "superior" to "other" clustering techniques, among others in that offers established goodness of fit criteria (e.g., BIC) and that it allows effectively test between n and m-cluster/class solutions.

A very readable intro into LCA is:

P. F. Lazarsfeld, Latent structure analysis, in: S. A. Stouffer, L. Guttman, E. A. Suchman, P. F. Lazarsfeld, S. A. Star, and J. A. Claussen, eds., Measurement and Prediction, (Wiley, New York, 1966).

A concise introduction is

Allan L. McCutcheon, 1987. Latent Class Analysis. Beverly Hills, CA: Sage Press.

Or peek into:

http://ourworld.compuserve.com/homepages/jsuebersax/

Once you are hooked, get:

Jacques A. Hagenaars and Allan L. McCutcheon (eds.) (2002) Applied Latent Class Analysis Models. Cambridge University Press.

AFAIK (v. 12), SPSS (just like SAS) does NOT support LCA (at least not as a single command). That is also the reason, why I think it is less popular in the social sciences.

You can get a good freeware program for LCA called lem, but that's a command line interface program. If you can afford LATENTGold, I would like to recommend it: It's a breeze to work with, when compared to lem or LCAG: It has a Windows driven interface, does not need start values (like some other LCA programs do), and it churns out results fairly quickly.

HTH,

Thomas

-- thomas koenig, ph.d. department of social sciences, loughborough university, u.k. http://www.lboro.ac.uk/research/mmethods/staff/thomas/index.html