Conditional Expectation Given a Sigma-Algebra · seed 1 · A4, ink-friendly. The answer key prints on its own page for grown-ups.
En anglais

Forecasting with what you know

Mathematics · Probability · ages 22-23
Name ______________________   Date ____________
  1. For integrable X and a sub-sigma-algebra G, every version of E[X|G] must be G-measurable.

    Circle one:   True   False

  2. Four outcomes a,b,c,d are equally likely. Let A={a,b,c}, B={b,c} and G be generated by {A,A^c}. What is E[1_B|G]?

    • 1/2 on A; 1/2 on A^c
    • 1 on A; 0 on A^c
    • 2/3 on A; 0 on A^c
    • 0 on A; 2/3 on A^c
  3. Four outcomes a,b,c,d are equally likely, with X-values 0,2,4,6. G has blocks {a,b} and {c,d}. The constant Y=3 has E[Y]=E[X], so it must be E[X|G].

    Circle one:   True   False

  4. Roll a fair six sided die. Given that the roll is even, what is the conditional expectation of the roll?

    • 3
    • 3.5
    • 4
  5. X takes -3,0,3 with probabilities 1/4,1/2,1/4. G records whether X=0. Since E[X|G]=0=E[X], X is independent of G.

    Circle one:   True   False

  6. For real X with E[X^2]<infinity, onto which space does E[X|G] project X in L2(F)?

    • The closed subspace of G-measurable L2 variables.
    • The collection of all G-events, viewed as subsets of Omega.
    • The subspace of constant variables, for every choice of G.
    • The space of all integrable F-measurable variables in L1.
  7. Four outcomes are equally likely. X has values 1,3,5,7. G groups the first two and last two outcomes. Z is 2 on the first block and -1 on the second. Compute E[(X-E[X|G])Z].

    Answer: ______________

  8. If integrable X is independent of the sub-sigma-algebra G, then E[X|G]=E[X] almost surely.

    Circle one:   True   False

  9. Two proposed versions of the same E[X|G], with integrable X, disagree on a set of probability 1/2. What follows?

    • Both satisfy the definition because versions may differ on positive-probability events.
    • At least one does not satisfy the conditional expectation definition.
    • G must be trivial because disagreement determines the information available.
  10. For every integrable X and any sub-sigma-algebras G,H, E[E[X|G]|H]=E[X|H] almost surely, even when H is not contained in G.

    Circle one:   True   False

LightMySky · lightmysky.comW1-mt_i2sxZc-ER8-s1

Answer key

For grown-ups. Fold this page away before handing over the rest.

Forecasting with what you know W1-mt_i2sxZc-ER8-s1

  1. True · G-measurability is required by the definition. The forecast can use only information in G.
  2. 2/3 on A; 0 on A^c · The indicator has values 0,1,1 on A, giving mean 2/3. On A^c={d}, it is 0.
  3. False · On A={a,b}, E[1_A X]=0.5 but E[1_A Y]=1.5. Matching overall means is insufficient.
  4. 4 · The even block holds 2, 4, and 6, whose local mean is 4.
  5. False · Both block means are 0, but given A={X=0}, the probability of X=0 is 1 instead of its unconditional probability 1/2.
  6. The closed subspace of G-measurable L2 variables. · The target is L2(G), the closed subspace of square-integrable G-measurable variables.
  7. 0 · The forecasts are 2 and 6. The residual times Z takes values -2,2,1,-1, whose mean is 0.
  8. True · The constant E[X] is G-measurable and satisfies the defining integral identity by independence.
  9. At least one does not satisfy the conditional expectation definition. · Versions of the same conditional expectation agree almost surely. A disagreement set of probability 1/2 is not null.
  10. False · For trivial G and H revealing a fair 0-or-1 variable X, the left side is 1/2 and the right side is X. Nesting cannot be omitted.
Worksheet · LightMySky