Realized power variation and stochastic volatility models (Q1395938)

Let \(y^*\) be a continuous semimartingale of the form \(y^*(t)=a^*(t)+\int_0^t\sigma(s) dw(s)\), where the adapted processes \(\sigma>0\) and \(a^*\) are assumed to be independent of the Brownian motion \(w\). For \(r\geq 1/2\), it is shown that, under some technical assumptions, the rescaled \(r\)-variation of \(y^*\) along an equidistant mesh converges to an integrated power volatility. More precisely, \[ \delta^{1-r/2}\sum_{i=1}^{\lfloor t/\delta\rfloor}|y^*(i\delta)-y^*((i-1)\delta)|^r\longrightarrow c\int_0^t|\sigma(s)|^r ds,\text{ in probability;} \] the constant \(c\) is equal to the \(r\)th moment of a standard normally distributed random variable. The corresponding central limit theorem is also proved.