Statistical Inference for Incomplete Ranking Data: The Case of Rank-Dependent Coarsening