Confidence intervals for the skew parameter of the skew normal distribution based on the Fisher transformation

MAMELI, VALENTINA;MUSIO, MONICA;Biggeri A.

2010-01-01

---

Abstract

In this work our aim is to estimate the distribution of the maximum between variables representing the natural logarithm of the PM10 emission in two stations of the town of Cagliari in Italy in 2004. It turns out that such order statistic has a skew normal distribution with skew parameter which can be expressed as a function of the correlation coefficient between the two initial variables. The skew-normal distribution belongs to a family of distributions which includes the normal one along with an extra parameter to regulate skewness. Azzalini (1985) was the first to introduce the skew-normal distribution and studied some of its properties. Loperfido (2002) showed that the distribution of the maximum between two standardized correlated normal variables, with correlation coefficient rho, is Skew-Normal with parameter lambda which depend of the correlation coefficient rho. In this specific case we show how is possible, using some theoretical results involving the correlation coefficient, to find an approximate confidence interval for the parameter of skew. This theoretical interval is then compared with parametric bootstrap intervals.