Abstract type for simulating cases

[SamuelBrand1]

Currently generating data, e.g. using data = missing patterns returns a Vector{Union{Missing, Real}} type object. This is known to be poorly performant because Real is an abstract type ( https://docs.julialang.org/en/v1/manual/performance-tips/#man-performance-abstract-container ).

This is not a priority issue, but worth flagging. A potential fix would probably involve the model looking at its error model, and making type specification e.g. PoissonError would specify a concrete integer type. That would be slightly awkward at the moment.

[seabbs]

Could you flesh this out a bit please sam? In particular what the performance implications are (given for most performance critical things you are passing data?)

[SamuelBrand1]

We no longer do this so closing