I reported this bug, maybe they'll fix it.
The whole domain is unusable to me. You can calculate relatively complex things if you type all parameters, but the moment you want to build few functions (to use them in later calculations) -- it all falls apart. For example in code below I wanted to build a function that will give me a distribution of sum of N independent variates with same distribution. Alas, it doesn't work for N > 1. Ok, I came up with a "fix" -- but that fix takes inordinate computation time if I use something more complex than uniform distribution -- most non-trivial computations simply time out after 1 minute.
(*simpled := DiscreteUniformDistribution[{1,6}]*)
simpled := BinomialDistribution[10,0.3]
transd := Module[{m,d},
TransformedDistribution[Max[m - d, 0], {m\[Distributed]simpled, d\[Distributed]simpled}]
]
sumd[n_Integer, dist_] := Module[{v,i,j,distlst},
distlst = Table[v[j] \[Distributed] dist,{j,n}];
TransformedDistribution[Sum[v[i],{i,n}],distlst]
]
fixd[dist_] := Module[{x,k,min,max},
(* unfortunately Quantile returns -inf for TransformedDistribution, so we'll use 0 for min (ok for our calcs) *)
{min,max} = {0, Quantile[dist, {0, 1}][[2]]};
EmpiricalDistribution[ Table[Probability[x==k, x\[Distributed]dist], {k,min,max}]->Table[k, {k, min, max}] ]
]
d0 := sumd[2, simpled]
PDF[d0]
d1 := sumd[1, transd]
PDF[d1]
(* this one is broken *)
d2 := sumd[2, transd]
PDF[d2]
d3 := sumd[2, fixd[transd]]
PDF[d3]
(* this one works but very slow with non-trivial simpled *)
d4 := sumd[6, fixd[transd]]
PDF[d4]
... and you can't use Set (as opposed to SetDelayed) with sumd because Table[...] refuses to evaluate with unknown parameter.
