Skew normal distribution explained
In probability theory and statistics, the skew normal distribution is a continuous probability distribution that generalises the normal distribution to allow for non-zero skewness.
Definition
Let
denote the
standard normal probability density function
}e^with the
cumulative distribution function given by
\Phi(x)=
\phi(t) dt=
\left[1+\operatorname{erf}\left(
}\right)\right],
where "erf" is the error function. Then the probability density function (pdf) of the skew-normal distribution with parameter
is given by
f(x)=2\phi(x)\Phi(\alphax).
This distribution was first introduced by O'Hagan and Leonard (1976).[1] Alternative forms to this distribution, with the corresponding quantile function, have been given by Ashour and Abdel-Hamid[2] and by Mudholkar and Hutson.[3]
A stochastic process that underpins the distribution was described by Andel, Netuka and Zvara (1984).[4] Both the distribution and its stochastic process underpinnings were consequences of the symmetry argument developed in Chan and Tong (1986),[5] which applies to multivariate cases beyond normality, e.g. skew multivariate t distribution and others. The distribution is a particular case of a general class of distributions with probability density functions of the form
where
is any
PDF symmetric about zero and
is any
CDF whose PDF is symmetric about zero.
[6] To add location and scale parameters to this, one makes the usual transform
. One can verify that the normal distribution is recovered when
, and that the absolute value of the
skewness increases as the absolute value of
increases. The distribution is right skewed if
and is left skewed if
. The probability density function with location
, scale
, and parameter
becomes
f(x)=
\right)\Phi\left(\alpha\left(
\right)\right).
The skewness (
) of the distribution is limited to slightly less than the interval
.
As has been shown,[7] the mode (maximum)
of the distribution is unique. For general
there is no analytic expression for
, but a quite accurate (numerical) approximation is:
Estimation
Maximum likelihood estimates for
,
, and
can be computed numerically, but no closed-form expression for the estimates is available unless
. In contrast, the
method of moments has a closed-form expression since the skewness equation can be inverted with
}
where
} and the sign of
is the same as the sign of
. Consequently,
},
\omega=
| \sigma |
\sqrt{1-2\delta2/\pi |
}, and
\xi=\mu-\omega\delta\sqrt{ | 2 |
\pi |
} where
and
are the mean and standard deviation. As long as the sample skewness
is not too large, these formulas provide method of moments estimates
,
, and
based on a sample's
,
, and
.
The maximum (theoretical) skewness is obtained by setting
in the skewness equation, giving
. However it is possible that the sample skewness is larger, and then
cannot be determined from these equations. When using the method of moments in an automatic fashion, for example to give starting values for maximum likelihood iteration, one should therefore let (for example)
|\hat{\gamma}1|=min(0.99,
| 3}|) |
|(1/n)\sum{((x | |
| i-\hat\mu)/\hat\sigma) |
.
Concern has been expressed about the impact of skew normal methods on the reliability of inferences based upon them.[8]
Related distributions
The exponentially modified normal distribution is another 3-parameter distribution that is a generalization of the normal distribution to skewed cases. The skew normal still has a normal-like tail in the direction of the skew, with a shorter tail in the other direction; that is, its density is asymptotically proportional to
for some positive
. Thus, in terms of the
seven states of randomness, it shows "proper mild randomness". In contrast, the exponentially modified normal has an exponential tail in the direction of the skew; its density is asymptotically proportional to
. In the same terms, it shows "borderline mild randomness".
Thus, the skew normal is useful for modeling skewed distributions which nevertheless have no more outliers than the normal, while the exponentially modified normal is useful for cases with an increased incidence of outliers in (just) one direction.
See also
External links
Notes and References
- O'Hagan. A.. Leonard. Tom. Thomas H. Leonard. 1976. Bayes estimation subject to uncertainty about parameter constraints. Biometrika. 63. 1. 201–203. 10.1093/biomet/63.1.201. 0006-3444.
- Ashour. Samir K.. Abdel-hameed. Mahmood A.. October 2010. Approximate skew normal distribution. Journal of Advanced Research. 1. 4. 341–350. 10.1016/j.jare.2010.06.004. 2090-1232. free.
- Mudholkar. Govind S.. Hutson. Alan D.. February 2000. The epsilon–skew–normal distribution for analyzing near-normal data. Journal of Statistical Planning and Inference. 83. 2. 291–309. 10.1016/s0378-3758(99)00096-8. 0378-3758.
- http://dml.cz/bitstream/handle/10338.dmlcz/124493/Kybernetika_20-1984-2_1.pdf Andel, J., Netuka, I. and Zvara, K. (1984) On threshold autoregressive processes. Kybernetika, 20, 89-106
- Chan. K. S.. Tong. H.. March 1986. A note on certain integral equations associated with non-linear time series analysis. Probability Theory and Related Fields. 73. 1. 153–158. 10.1007/bf01845999. 121106515. 0178-8051. free.
- Azzalini . A. . 1985 . A class of distributions which includes the normal ones. Scandinavian Journal of Statistics . 12 . 171–178.
- Book: Adelchi. Azzalini. Antonella. Capitanio . 2014 . The skew-normal and related families. 978-1-107-02927-9 . 32–33.
- http://www.tandfonline.com/doi/pdf/10.1080/02664760050120542 Pewsey, Arthur. "Problems of inference for Azzalini's skewnormal distribution." Journal of Applied Statistics 27.7 (2000): 859-870