This thesis concerns itself with the statistical variation in STR NGS data, with application in forensic genetics. We introduce simple methods for DNA profiling in single contributor samples, and afterwards examine the quality associated with NGS reads. The errors are examined, first the systematic errors, stutters and shoulders, and then the more general noise. The general noise is handled using a noise threshold, which imposes drop-outs in the data. The heterozygote imbalance is therefore examined and a model for full coverage is presented. Thereafter, the probability of dropout is predicted and the thesis concluded.