I have asked this on cross validated, but I will also ask it herer.
I have a set of $2$-tuples:
$$\{(x_1,y_1),(x_2,y_2),...,(x_{500},y_{500})\}$$
Where every $x_i$ is the volatility and every $y_i$ is the log return. So I am trying to gauge the link between the volatility and the returns. There is a bump in the observations for $y_i < 0.1$ (for volatilities less than $10\%$). These observations are based off a trading strategy. I am trying to understand the statistical significance of this bump. I decided to use bootstrapping, measure the skewness of observations for volatilitiles less than $10\%$ in each case, construct a cofidence interval and then see whether the original skewness falls in that confidence interval. But the question is, how do I construct a skewness measure for $2$-tupels. Here is the depiction of my results:
and here is the cross validated question.
EDIT: Ahh! Perhaps the distances of the observations of positive log returns to the horizontal axis at $0$? But that would sum the small ones as well. I suppose I would like to ignore them..