A preponderance of zeros may occur with several types of data. With interval data one may have a large proportion of actual zero values. For example, if income is the response being modeled, there could be many subjects with no income. With ordinal data one may be faced with the problem that most subjects have all zero values. If adverse event data is being modeled and 0 reflects no adverse event, then one might have a large fraction of subjects with all zero values. PRI in conjunction with several pharmaceutical companies has been exploring ways to model such responses. These include altered zero, added zero, mover-stayer, and sojourn models to name a few.
