Brownian typeclass vs struct

Stream: Brownian motion

[Joris van Winden]

I am a bit confused about the choice of typeclass vs structure/def with regards to probability theory. I found the following exampes:

Typeclass: IsPreBrownian, IsBrownian, IsGaussian
Structure/def: HasLaw, IsGaussianProcess, HasIndepIncrements

Is there a guiding principle behind why some of these are typeclasses and some are not?

[Etienne Marion]

IsGaussian (which is the only one which already existed before the project) is a typeclass because it allows to automatically infer IsProbabilityMeasure. The other typeclasses are about random variables and I think they should actually be structures because they are often too far fetched to be inferred automatically. For instance when I upstreamed HasGaussianLaw and IsGaussianProcess I turned them into structures while they were originally classes.