Latent Class for Multinomial Logit (MNL) is a popular procedure for finding segments of respondents with different preferences from choice data such as CBC and MaxDiff. However, one aspect of standard Latent Class analysis that may interfere with some analysts’ goals is that it sometimes can form segments that mainly differ in terms of scale (response error) but don’t differ much in terms of real preference patterns.
Bryan Orme proposes a simple method for constraining latent class solutions, assuming that the standard deviation across a utility vector represents a proxy for scale. He demonstrates using an artificial data set that scale constrained latent class avoids distinguishing between high-error and low-error respondents who otherwise have identical preferences.
The scale constraints described within this white paper are available as an option within Sawtooth Software’s latent class modules starting with the SSI Web v8.3 release.