You can do this, but first of all, you should ask yourself a question: Why to play with exponent at all, and why to apply different exponents to different groups. Do you have a really good market data for different submarkets? Do you expect that your survey data should mimic it? And why those submarkets should have different scale parameters? I'ts much easier and often safer not to change exponent. Anyway, in custom built simulators segment specific scale parameters are trivial (just multiply selected subgroups utilities by whatever you chose). In the SMRT (at least in my version) it's harder, you have to edit hbu files or conduct several simulations and aggregate you result.