Skip to main content
1 of 2
zhoraster
  • 1.5k
  • 11
  • 24

Ok, here is a counterexample with the index set $\{1,2\}$ (you can easily extend it to whole $\mathbb N$ if you wish).

Let $(Y_1,Y_2)$ be independent Bernoulli($1/2$) and set $(X_1,X_2) = (Y_1,Y_2)$ if $Y_1 + Y_2 > 0$; if $Y_1=Y_2 = 0$, set $X_{1} = 0$, $X_{2} = 1$.

In this case $$ \mathrm P(X_1 = 1\mid X_2 = 1) = \frac{P(X_1 = 1, X_2 = 1)}{P(X_2 = 1)} = \frac{1/4}{3/4} = \frac13<\frac12. $$

zhoraster
  • 1.5k
  • 11
  • 24