Personally, I prefer to use more of a "lumped mass" model where I lump the mass at each major floor level. I feel like that is an easier / cleaner way to analyze a large structure.
If we're interested in local vibration in areas (like for machinery or such) then using self weight of elements along the height of the wall makes sense. But, when we're talking about the main lateral force resisting system for earthquake analysis, I feel like letting the mass distribute out only increases the analysis complexity without adding much accuracy.
Others may have different preferences, but this is mine.
One of my first assignments as a structural engineer was to do a dynamic analysis of a nuclear plant building. My model consisted of 4 sticks and 4 masses. Years later someone did an analysis with ANSYS with 10's of thousands of solid elements. Global results were pretty much the same. It's amazing how many nuclear plants were designed in the early days with simple stick models.
