Aggregating preferences in multi-issue domains by using maximum likelihood estimators