In our model, we have used fully conjugate prior distributions throughout enabling efficient analytical integration to be performed for posterior calculations. Estimates for hyperparameters in prior distributions were obtained by learning from reference datasets of chromosome X multiple copy cell lines (20), for which the copy numbers are known, using maximum marginal likelihood inference. For the remaining model parameters, we have chosen to set these ‘objectively’ in order to calibrate our model to user-specified false positive error rates.
