Fitting Unreplicated Choice Experiments