# Expected value of fractional part is less than 1/2

(My apologies in advance; this is very open-ended but I ask leave to post regardless.)

I'm trying to recall a theorem on the fractional part of... some fairly natural class of sequences. It showed that the expected value is not 1/2, as might be assumed, but rather some smaller value (perhaps around 0.4). Unfortunately I can't think of what sorts of sequences these were, and that makes it quite hard to recall the theorem itself.

It was not about some contrived sequence like the Pisot/PV numbers. If I can think of additional details I will edit them in or add them as a comment.

-
Was it related to simple continued fractions? – Douglas Zare Apr 6 '11 at 5:57

Here is another: If $X$ is uniform on $[0,1]$ then $\frac {1}{X} - \lfloor \frac{1}{X} \rfloor$ is not uniform on $[0,1]$. For example, the density at $1/2$ is $\frac{4}{9} + \frac{4}{25} + \frac{4}{49} + ... = \frac{\pi^2}{2} - 4 \approx 0.9348.$ The expected fractional part of $1/X$ is $1-\gamma \approx 0.422784.$