Olsen and Schaffer (2001) proposed a growth model for semicontinuous outcomes (e.g., alcohol use) that have a substantial portion of identical values, typically zero (e.g., nonusers), and then a continuous and typically skewed distribution for the rest of the values (representing variations in drinking levels among users). Part I is a growth model for a binary variable that indicates any alcohol use versus nonuse. Part II is a standard growth model for, in the current study, volume of alcohol use given any use at all, and individuals only have a non-missing score at a particular assessment if the alcohol score is above zero.
