No. In that case, I'd recommend using HB to estimate the utilities. Export the HB utilities (via SMRT + Run Manager + Export) as zero-centered diffs (to normalize the utilities across respondents). Then, drop the levels that shouldn't have an influence in the clustering. Then, re-zero center the attributes that had levels dropped. Then, submit those utilities to our CCEA package for cluster ensemble analysis (better than k-means clustering!).