Motivated by the need to generate high-quality synthetic data, we sought to develop efficient methods for drawing large numbers of independent samples from conditioned probability distributions. For this, we developed and evaluated two novel methods to sample from a probabilistic graphical model, where the model is compiled to an arithmetic circuit. The methods are a form of inverse transform sampling. Our contribution is the decomposition of a cumulative probability distribution into a sum of a small number of terms, where each term can be efficiently calculated using the arithmetic circuit. For small models, the proposed methods dominate. For larger models, dependent samplers exhibited faster performance. However, within the context of independent samplers, our proposed method remained the most efficient.