Book And Research Paper Forum • Re: End-of-year Competition: David Orrell's q-variance paper in Wilmott

Do you mean this? Q: Figure 1 shows binned variance versus z. Why not just do a linear regression of variance versus z2 to estimate the quadratic coefficient? A: Q-variance is a statement about the conditional expectation function E[v(T)∣z], in other words the expected variance given a value of z. Ordinary least squares regression of v on z2 does not test this claim directly; rather, it estimates the best linear predictor of v in terms of z2, characterized by orthogonality conditions. Nonparamet