I am currently trying to calculate some expressions of means of iid (independent and identically distributed) random variables. I try to verify my manual calculations with Mathematica: I defined

m[Subscript[X, i_]] := m[Subscript[X, 1]] /; i != 1
m[Subscript[X, i_] Subscript[X, j_]] := m[Subscript[X, i]] m[Subscript[X, j]] /; i != j

which works fine. ($E(X_j)=E(X_1)$ and $E(X_iX_j)=E(X_i)E(X_j)$ if $i\neq j$)

I know that Mathematica can simplify sums with variable bound like Sum[a, {i, 1, n}] to an.

Now I consider a sum for a fixed $j$:

Sum[m[Subscript[X, i] Subscript[X, j]], {i, 1, n}]

Is it possible to make Mathematica simplify it to $(n-1)E(X_1)^2+E(X_j^2)$?


This isn't as general as one would like but it might be a start.

First I would drop the subscripts (Subscript[X,i]) and use indices (X[i]). (There are numerous references at this site explaining why.)

Then I would define the potential outcomes of expectations of powers and products of the independent and identically distributed random variables (and, of course, assuming that all of the appropriate moments exist).

m[X[i_]^k_] := μ[k]
m[X[i_] X[j_]] := μ[1]^2 /; i != j
m[X[i_]^k_ X[j_]] := μ[k] μ[1] /; i != j
m[X[i_] X[j_]^k_] := μ[1] μ[k] /; i != j
m[X[i_]^ki_ X[j_]^kj_] := μ[ki] μ[kj] /; i != j
m[X[i_]] := μ[1]

(Note that this only defines the product of two random variables. Other definitions - or rules - would need to be added for the expectation of 3 or more random variables.)

Then one can evaluate the sum of interest by fixing n and j:

n = 20
j = 2
Sum[m[X[i] X[j]], {i, 1, n}]
(* 19 μ[1]^2+μ[2] *)

I'm sure there must be ways to allow for general n but I don't know enough to be able to do that.


It's probably more straightforward to apply the expectation function as a rule:

n = 20
j = 2
Sum[X[i] X[j], {i, 1, n}] /. {X[i_]^k_ -> μ[k], X[i_] -> μ[1]}
(* 19 μ[1]^2 + μ[2] *)
| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.