Conclusions on the development of delinquent behaviours during the life-course can only be made by analysing longitudinal data. Naturally, these data are affected by missingness on the unit level (i.e. not all participants were available at all panel waves and some even totally dropped out of the study). Reinecke and Weins (2013) have analyzed data from an adolescents’ four-wave panel (for details, see www.crimoc.org) using full-information maximum likelihood estimation and normal-model multiple imputation. Unfortunately, delinquency data (such as the count of delinquent behaviours in the previous year) are seldomly normally distributed. In fact, the destribution is usually quite highly skewed with an excess of zero counts. Since then, more appropriate imputation methods and models have become available such as package countimp by Kleinke and Reinecke (2019). In the present paper, we introduce version 2 of package countimp, re-analyze the data by Reinecke and Weins (2013), and compare results of their zero-inflation growth curve model based on different imputation methods with appropriate and not so appropriate distributional assumptions. We conclude by encouraging applied researchers to choose multiple imputation methods with fitting distributional assumptions.