construction, weak convergence
Let dμn be a sequence of probability measures on [0,1] so that
∫f(x)dμn(x)
converges for every continuous function f:[0,1]→R.
-
Show that
∬g(x,y)dμn(x)dμn(y)
converges for every continuous g:[0,1]2→R.
-
Show by example that under the above hypotheses, it is possible that
∬{0≤x≤y≤1}dμn(x)dμn(y)
does not converge.
Solution.
-
By Stone-Weierstrass, the set
A={k=1∑nfk(x)gk(y)∣∣n∈N, fk,gk∈C([0,1])}
is dense in C([0,1]2). Hence, by applying weak convergence and Fubini's theorem, we see that μn⊗μn converges weakly-* to μ⊗μ.
-
Let μn=δ0 if n is odd and μn=n1∑k=1nδk/n2 for n even. Then for n even and g is continuous,
∫g(x)dμn(x)=n1k=1∑ng(n2k)n→∞g(0),
by continuity, and it's clear that the same holds for the odd terms. However, for n od, we have
∫{0≤x≤y≤1}dμn(x)dμn(y)=0
and for n even,
∫{0≤x≤y≤1}dμn(x)dμn(y)=∫01∫0ydμn(x)dμn(y)=j=1∑n2∫[n2j−1,n2j)∫0ydμn(x)dμn(y)=j=1∑n∫[n2j−1,n2j)njdμn(y)+j=n+1∑n2∫[n2j−1,n2j)nndμn(y)=j=1∑nn2j=2nn+1n→∞21,
so the sequence diverges.