Interpreting correlated random parameters in choice experiments