Documente Academic
Documente Profesional
Documente Cultură
Rolf Poulsen
November 17, 2010
In general:
E(f (Y )) 6= f (E(Y ))
If f linear, then equality holds. Otherwise, you are ill advised to assume that
it does. In case f is convex, we have E(f (Y )) ≥ f (E(Y )) with strict inequality
unless in degenerate cases. This is known as Jensen’s Inequality.
2 /2
Theorem 1: If Z ∼ N (0, σ 2 ) then E(eZ ) = eσ .
1
The term in the square bracket looks nasty, but upon closer inspection is just
the integral on the density of a N (σ 2 , σ 2 )-variable, and hence a simply a con-
voluted way of writing 1 — and we’re done.
2 /2
Corollary: If X ∼ N (µ, σ 2 ) then E(eX ) = eµ+σ .
2 /2
Theorem 2 : If Z ∼ LN (µ, σ 2 ) then E(Z) = eµ+σ and
2 2
Var(Z) = e2µ+σ (eσ − 1).
Proof: We may write Z = eX , where X ∼ N (µ, σ 2 ). The part first then follows
directly from the Corollary. To calculate the variance, let’s first write
The final expression follows from the definition Var(Z) = E(Z 2 ) − [E(Z)]2 .
The results in Theorem 2 are used repeatedly and “without much warning” in
the Examiners’ Reports for CT1 exams (i.e. in the official solutions.)