Latent class analysis (LCA) is an analogue to factor analysis for discrete data. It has received attention for its usefulness in validating diagnostic categories in the absence of a gold standard. LCA may be equally useful in exploring uncharacterized prognostic groups of patients in clinical settings.
Until June, 2006, I was unaware of LCA being available in SAS, a statistical software that is widely used in public health and medicine. In March 2006, I presented a paper (immediately below) that describes an approach to latent class analysis in SAS. I was gratified by the response, especially among researchers in marketing, who frequently use LCA and are very interested in its integration with the widely-used SAS System.
This version of the program (posted May 30, 2006) should run by itself without the need for additional information or data. It requires that your data be structured like this sample data set, after you've converted it into a SAS dataset. The program's current version detects two latent classes on the basis of observed measures of four binary manifest indicators (in the variables a,b,c, and d). The data format includes counts (in the variable FREQ) for each of sixteen possible response profiles.