I am self-learning about DCT and MCT and related Lemma's. I understand the theorems as constructed, but I am struggling to apply them.

As an example:

$X_n=|Z|^{1/n}$ where $Z\sim N(0,1)$

I am trying to do the following:

a)Identify $X$ such that $X_n\rightarrow X$ a.s

b)Determine if $E(X_n)\rightarrow E(X)$ as $n\rightarrow\infty$

I don't think the answer requires exact calculation of expectation, but seeing that done explicitly might help me get my footing.

Work/attempt: This is not a homework question, so any level of detail is extremely appreciated and welcomed.

This is what I know of MCT/DCT:

Monotone Convergence: $f_1,f_2...f_n\uparrow f$ as $n\rightarrow\infty\Rightarrow \lim_{n\rightarrow\infty}\int f_nd\mu=\int fd\mu$

Dominated Convergence: Given measurable functions $f_1,f_2...$ and $|f_n|\le g$ for some $g$, and if $f_n\rightarrow f \mu-a.e.$, then $\lim_{n\rightarrow\infty}\int|f_n-f|d\mu=0$ and $\lim_{n\rightarrow\infty}\int f_nd\mu=\int fd\mu$

So in my example $X_n$ is a series of nth roots of standard normal RVs. I'm not seeing any monotone convergence, so I look to DCT.

Now the trick is to find a function, $g$, that is always greater than or equal to my series. I'm not sure what $X$ this converges to, which makes it hard to find $g$. My first guess is something like $g=|Z|^n$, but I have no theoretical grounding for this choice, and it's not clear to me that it even meets my criteria.


  • $\begingroup$ In the DCT, one tries to get $|f_n|\leqslant g$ with $g$ integrable and independent of $n$. Hence $g=|Z|^n$ cannot do anyway. $\endgroup$ – Did Jun 22 '12 at 5:45

a) $X=\mathbf 1_{Z\ne0}$ (consider separately the cases $Z(\omega)=0$ and $Z(\omega)\ne0$ and proceed).

b) Since $X_n\leqslant1+|Z|$ for every $n\geqslant1$ (consider separately the cases $|Z(\omega)|\lt1$ and $|Z(\omega)|\geqslant1$ and proceed), Lebesgue dominated convergence theorem implies that $\mathrm E(X_n)\to\mathrm E(X)=\mathrm P(Z\ne0)=1$.

Note: One can also apply Lebesgue monotone convergence theorem twice, separately to the random variables $U_n=|Z|^{1/n}\,\mathbf 1_{|Z|\leqslant1}$ and $V_n=|Z|^{1/n}\,\mathbf 1_{|Z|\gt1}$. To see this, note that $X_n=U_n+V_n$ for every $n\geqslant1$, the sequence $(U_n)$ is nondecreasing, the sequence $(V_n)$ is nonincreasing, $U_n\to U=\mathbf 1_{0\lt|Z|\leqslant1}$ and $V_n\to V=\mathbf 1_{|Z|\gt1}$. Hence $\mathrm E(X_n)\to\mathrm E(U)+\mathrm E(V)=1$.

| cite | 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.