Lets say I construct a parametrisation of the volatility surface that lets me infer dynamics i.e correlation between strike vol.
Is computing the sample correlation (after controlling for spot-vol correlation) a reasonable way to test this model? Is there a better framework to answer this question?
