Amos 7.0 can estimate posterior predictive distributions for missing data values and for scores on unobserved numeric variables that underlie ordered-categorical and censored measurements.

In a latent variable model, there are three types of unknown numeric values:

1. Parameter values and functions of parameter values, for example regression weights and correlations

2. Missing data values

3. Partially missing data values such as ordered-categorical or censored measurements

In a Bayesian analysis these three types of unknowns are all treated in the same way. The state of knowledge about any unknown quantity is represented by a posterior density that shows which values are probable. In the case of data values that are missing or partially missing the posterior density is called a posterior predictive distribution.