Counting the number of parasite eggs in faecal samples is a widely used diagnostic method to evaluate parasite burden. Typically a sub-sample of the diluted faeces is examined for eggs. The resulting egg counts are multiplied by a specific correction factor to estimate the mean parasite burden. To detect anthelmintic resistance, the mean parasite burden from treated and untreated animals are compared. However, this standard method has some limitations. In particular, the analysis of repeated samples may produce quite variable results as the sampling variability due to the counting technique is ignored. We propose a hierarchical model that takes this sampling variability as well as between-animal variation into account. Bayesian inference is done via Markov chain Monte Carlo sampling. The performance of the hierarchical model is illustrated by a re-analysis of faecal egg count data from a Swedish study assessing the anthelmintic resistance of nematode parasite in sheep. A simulation study shows that the hierarchical model provides better classification of anthelmintic resistance compared to the standard method.