Temper tantrums are a normal part of growing up in young children and they can be very frustrating for parents to handle. Good knowing 
of tantrums can help parents be in control of the situation and to predict children's future antisocial behaviors. The two main features of 
tantrum are duration and frequency. In this project, the analysis focused on tantrum temporal patterns, which exist within one day, as well 
as across three weeks. The data came from the telephone interviews of parents of 18 through 60 month old children in Wisconsin. Statistical 
analysis showed that the tantrum frequency within a day was fit to a bimodal distribution and the tantrum frequency across three weeks 
followed a normal mixture model with 3 components. There was no circadian rhythm in tantrum duration within a day as well as across three 
weeks. Determination of best model fit among the unimodal, bimodal and multimodal distributions was based on Bayesian Information 
Criterion (BIC). This analysis was carried out using the MCLUST package on R. We also concluded that most tantrum durations can not 
be predicted based on previous tantrum durations by using time-series autoregression.