Journal of Modern Applied Statistical Methods Volume 5 Issue Article 3 --005 JMASM5: Computing Percentiles of Skew- Normal Distributions Sikha Bagui The University of West Florida, Pensacola, bagui@uwf.edu Subhash Bagui The University of West Florida, Pensacola, sbagui@uwf.edu Follow this and additional works at: http://digitalcommons.wayne.edu/jmasm Part of the Applied Statistics Commons, Social and Behavioral Sciences Commons, and the Statistical Theory Commons Recommended Citation Bagui, Sikha and Bagui, Subhash (005) "JMASM5: Computing Percentiles of Skew-Normal Distributions," Journal of Modern Applied Statistical Methods: Vol. 5 : Iss., Article 3. DOI: 0.37/jmasm/6355400 Available at: http://digitalcommons.wayne.edu/jmasm/vol5/iss/3 This Algorithms and Code is brought to you for free and open access by the Open Access Journals at DigitalCommons@WayneState. It has been accepted for inclusion in Journal of Modern Applied Statistical Methods by an authorized editor of DigitalCommons@WayneState.
Journal of Modern Applied Statistical Methods Copyright 006 JMASM, Inc. November, 006, Vol. 5, No., 575-588 538 947/06/$95.00 JMASM5: Computing Percentiles of Skew-Normal Distributions Sikha Bagui Subhash Bagui The University of West Florida An algorithm and code is provided for computing percentiles of skew-normal distributions with parameter λ using Monte Carlo methods. A critical values table was created for various parameter values of λ at various probability levels of α. The table will be useful to practitioners as it is not available in the literature. Key words: Skew normal distribution, critical values, visual basic. Introduction The skew normal ( SN) class of densities is a family of densities that includes the normal density, with an extra parameter than the normal one to regulate skewness. In other words, SN class of densities extends the normal model by allowing an extra shape parameter to account for skewness. Let Z be a standard normal random variable with probability density function (pdf). φ and cumulative distribution function (cdf) Φ. A random variable Z is said to have the oneparameter skew-normal distribution with parameter λ, (say, Z ~ SN( λ )), if it has the following density function f (; z λ) = () φ z Φ ( λz), < z < () Z where λ is a real valued parameter. When λ = 0, one gets the standard normal density, for λ > 0, one gets the positively skewed Sikha Bagui is an Assistant Professor in the Department of Computer Science. Her areas of research are database and database design, data mining, pattern recognition, and statistical computing. email: bagui@uwf.edu. Subhash Bagui is Professor of Statistics at the University of West Florida. His areas of research are statistical classification and pattern recognition, bio-statistics, construction of designs, tolerance regions, and statistical computing. email:sbagui@uwf.edu distribution, and for λ < 0, one gets the negatively skewed distribution. Graphs of skewnormal densities for different values of λ are shown in Figure. -0-5 0 5 0 Figure : Skew-Normal Plot with λ = 4,, 0,, 4 The skew-normal distributions first indirectly appeared in Roberts (966), O Hagan and Leonard (976), and Aigner and Lovell (977). The landmark article by Azzalini (985, 986) gave a systematic treatment of this distribution and studied its fundamental properties. Recent work on characterization of skew-normal distribution can be found in Arnold and Lin (004) and Gupta, Nguyen, and Sanqui (004a). Azzalini and Dalla (996), Azzalini and Capitanio (999), and Gupta, Gonzalez- Farias, and Dominguez-Molina (004b) investigated the multivariate extensions of this distribution. There is always a growing interest in modeling non-normal distributions. This 575
576 PERCENTILES OF SKEW-NORMAL DISTRIBUTIONS model can be very useful to capture nongaussian behavior of the data. It already proved to be useful in modeling real data sets (Azzalini & Captanio, 999). To capture non-gaussian behavior of the data, Aigner and Lovell (977) used this density in stochastic frontier models, Adcock and Shutes (999) used it in portfolio selection of financial assets, and Bartolucci, De Luca, and Loperfido (000) used it in detection of skewness in stock returns. Below some important properties of SN( λ ) are stated without proof. Let and δ = λ + λ z Φ ( z) = f ( t; λ) dt. λ Z. SN(0) has the density N (0,).. If Z ~ SN( λ ), then Z ~ SN( λ). 3. fz (; z λ) is strongly unimodal, i.e. log fz (; z λ) is a concave function of z. 4. Φλ ( z) = Φ λ ( z), i.e. fz (; z λ) is a mirror image of f (; z λ). χ (chi- 5. If Z ~ SN( λ ), then square with one d.f.). Z Z ~ 6. If Z ~ SN( λ ), then Z has the m.g.f. M Z t () t = exp( ) Φ ( δt). 7. If Z ~ SN( λ ), then EZ ( ) = and Var ( Z ) =. π δ π δ 9. Let Z and Z be independent N (0,) random variables. Set Z if Z < λz Z = Z if Z > λz. Then Z ~ SN( λ ). 0. Let ( Z, Z ) be a bivariate normal with EZ ( ) = EZ ( ) = 0,Var ( Z ) = Var ( Z ) =, and corr ( Z, Z) = ρ. Set Z = max( Z, Z ). Then Z ~ SN( λ ), ρ + ρ where λ =. Property (4) tells us that fz (; z λ) is a mirror image of fz (; z λ). So it is sufficient to find percentiles only for positive values of λ. For negative values of λ, one may use property (4) to derive the percentiles. In order to simulate skew-normal random variate one may use either properties (8), or (9), or (0). It is also important that sufficient number of random variates must be used to model the distribution of Z adequately. Below we extend one parameter skew-normal distribution to three parameters skew-normal distributions. Definition. (Three-parameter skew-normal). A random variable X is said to have the threeparameter skew-normal distribution (denoted by X ~ SN( μ, σ, λ )) if X = μ + σ Z, where Z ~ SN( λ ), i.e. if X has the following density function fx λ = φ Φ λ, < x < x μ x μ ( x; ) σ ( σ ) ( σ ) () 8. Let Z and Z be independent N (0,) random variables, then Z = δ Z + δ Z ~ SN( λ ).
BAGUI & BAGUI 577 Methodology Property (8) will be used in order to generate skew-normal random variate Z. Before this property is used, standard normal variates must be simulated. The following theorem explains how to generate standard normal variates. Theorem. (Box & Muller, 958). Let U and U be independent random variates from uniform (0,) ( i..e. U, U ~ U (0,) ), then the variates ( ln ) sin π ( ln ) cos π Z = U U ; Z = U U are independent standard normal variates. Proof. The joint density of U and U is given by f ( u, u ) =. Also, from the above U, U transformation note that u z π z ( z z ) u = e + and = tan. Thus, the Jacobian of this ( z + z ) transformation is given by J = π e. Hence, the joint density of Z and Z is given by f ( z, z ) = Z, Z π ( z + z ) = e. f ( u, u ) J U, U (3) The equation (3) confirms that Z and Z is independent standard normal variates. Now using the above theorem and the property (8), the skew-normal variates Z shall be generated. The algorithm is as follows. Algorithm λ. Define δ = + λ. Generate two independent random variates U and U from U (0,). 3. Set ( ln ) cos π Z U U and ( ln ) sin π Z U U. 4. Deliver Z = δ Z + δ Z. This process is repeated until the density of Z is adequately modeled. The code is written in Visual Basic. After a sufficient number of runs the program will return percentiles corresponding to the probability level α = 0.0, 0.0, 0.05, 0.05, 0., 0.99, 0.095, 0.975, 0.98, 0.99 for a given λ value. For λ = 0.0 to 50, in Table A, the percentiles of Z are provided for α = 0.0, 0.0, 0.05, 0.05, 0.; and in Table B, the percentiles are provided for α = 0.99, 0.095, 0.975, 0.98, 0.99. Let the notation Z α ( λ) denote the α (00)% percentile of Z for a given value of λ. Percentiles for negative λ : Let λ > 0, then Z α ( λ) can be derived by means of ( λ) = ( λ), as fz Z α Z α (; z λ) is a mirror image of f (; z λ). Percentiles of X : For the three-parameter skew normal r.v. X ~ SN( μ, σ, λ ), let X α ( λ) denote the α (00)% percentile of X for a given value of λ. Since X = μ + σ Z is a monotonically increasing transformation, the percentile X α ( λ) of X can be obtained from the equation Z X ( λ) = μ + σz ( λ). (4) α α
578 PERCENTILES OF SKEW-NORMAL DISTRIBUTIONS Table A. Percentiles of Skew-normal distributions λ α 0.0 0.0 0.05 0.04 0.05 0.0 0.0 -.3660 -.043600 -.949637 -.74096 -.63494 -.70585 0.0 -.30865 -.03530 -.94678 -.7343 -.6663 -.6473 0.03 -.99909 -.0709 -.93309 -.73888 -.6798 -.5435 0.04 -.974 -.08600 -.94809 -.7555 -.60959 -.4603 0.05 -.890 -.000 -.9639 -.70700 -.6078 -.3776 0.06 -.73634 -.00495 -.90778 -.698454 -.59648 -.93 0.07 -.64764 -.9963 -.89906 -.689889 -.584055 -.095 0.08 -.55799 -.98383 -.890096 -.68037 -.575384 -.358 0.09 -.46697 -.974694 -.885 -.6753 -.566573 -.03799 0.0 -.3753 -.965540 -.873 -.66374 -.55778 -.9568 0.5 -.8965 -.9859 -.8530 -.674 -.577 -.556 0.0 -.37967 -.86908 -.776368 -.569685 -.4657 -.0655 0.5 -.084749 -.87640 -.75684 -.5065 -.4685 -.060535 0.30 -.09456 -.764838 -.673795 -.47057 -.367758 -.04637 0.35 -.97884 -.76 -.60986 -.40003 -.3857-0.968773 0.40 -.95085 -.656844 -.567945 -.369303 -.6877-0.93367 λ α 0.0 0.0 0.05 0.04 0.05 0.0 0.45 -.85759 -.60308 -.5466 -.38676 -.9467-0.878647 0.50 -.7994 -.54898 -.4683 -.687 -.7097-0.834859 0.55 -.74834 -.49488 -.409644 -.9545 -.354-0.7900 0.60 -.68486 -.44066 -.358373 -.74 -.07677-0.750845 0.65 -.699 -.390497 -.30856 -.4578 -.03460-0.7095 0.70 -.57436 -.3400 -.59576 -.079053-0.987594-0.676 0.75 -.560 -.9533 -.9 -.035095-0.94533-0.635839 0.80 -.469659 -.4474 -.6668-0.99756-0.904630-0.600690 0.85 -.4968 -.98740 -.574-0.95997-0.86553-0.5674 0.90 -.3767 -.54904 -.08078-0.9938-0.8805-0.53587 0.95 -.3488 -.776 -.03948-0.875538-0.799-0.504804.0 -.8050 -.07379 -.000580-0.839787-0.75858-0.475989. -.9590-0.996480-0.97569-0.77304-0.694478-0.4673. -.8556-0.97-0.8609-0.7349-0.636756-0.374875.3 -.047759-0.86390-0.80043-0.65786-0.58459-0.33973.4-0.98884-0.8065-0.744969-0.607346-0.537-0.9356.5-0.9373-0.753687-0.694679-0.5604-0.494470-0.59078.6-0.86950-0.705780-0.6489-0.50-0.45575-0.84.7-0.89759-0.66996-0.60735-0.48366-0.40568-0.0038.8-0.77430-0.6985-0.56908-0.449680-0.388656-0.753.9-0.73466-0.58538-0.53444-0.4868-0.35960-0.56.0-0.69404-0.55756-0.509-0.390393-0.333-0.353. -0.658558-0.50897-0.47897-0.364488-0.30899-0.3589. -0.65889-0.49486-0.445939-0.34075-0.86783-0.096753.3-0.59567-0.46686-0.47-0.389-0.6644-0.08409.4-0.567697-0.44084-0.398-0.9884-0.47777-0.067438.5-0.54678-0.49685-0.376990-0.80305-0.0357-0.05467.6-0.57496-0.398873-0.357355-0.6387-0.4697-0.0498.7-0.49507-0.379556-0.33906-0.4730-0.0000-0.037.8-0.474034-0.3658-0.37-0.3604-0.8649-0.096
BAGUI & BAGUI 579 Table A. Continued.9-0.45445-0.344773-0.30639-0.893-0.7380-0.039 3.0-0.436086-0.39099-0.9554-0.0690-0.609-0.004783 3. -0.4885-0.3440-0.77760-0.9433-0.560 0.0098 3. -0.40685-0.300694-0.64865-0.833-0.40979 0.0044 3.3-0.387500-0.878-0.5750-0.786-0.348 0.0688 3.4-0.3739-0.75705-0.4405-0.636-0.593 0.0994 3.5-0.359675-0.6495-0.3070-0.5404-0.45 0.0874 3.6-0.346930-0.53540-0.0636-0.4547-0.06434 0.03409 3.7-0.3349-0.4346-0.50-0.3740-0.099090 0.039075 3.8-0.33507-0.33839-0.04-0.9843-0.0900 0.04377 3.9-0.370-0.480-0.93740-0.698-0.085694 0.048080 4.0-0.30460-0.63-0.85747-0.5946-0.07957 0.0550 4. -0.9758-0.0806-0.788-0.09578-0.073800 0.055953 4. -0.83507-0.00385-0.70970-0.0353-0.068333 0.059509 4.3-0.74730-0.93095-0.6476-0.097830-0.06378 0.06874 4.4-0.66349-0.864-0.5769-0.09408-0.05888 0.066000 4.5-0.58354-0.79500-0.558-0.08770-0.05365 0.068956 4.6-0.5070-0.7380-0.45664-0.08397-0.04965 0.07733 4.7-0.43464-0.6745-0.4008-0.07775-0.045097 0.074366 4.8-0.3657-0.649-0.34749-0.073343-0.044 0.076857 4.9-0.9873-0.5594-0.9676-0.06935-0.037373 0.07959 5.0-0.3494-0.50707-0.487-0.06539-0.033789 0.08360 λ α 0.0 0.0 0.05 0.04 0.05 0.0 5. -0.7406-0.4578-0.08-0.0637-0.03039 0.083439 5. -0.567-0.4094-0.5759-0.057684-0.0737 0.085408 5.3-0.05935-0.363-0.59-0.05498-0.04034 0.08769 5.4-0.00537-0.3905-0.0747-0.050879-0.007 0.08906 5.5-0.95384-0.770-0.0354-0.047696-0.0856 0.090667 5.6-0.90408-0.3636-0.099800-0.044650-0.05548 0.096 5.7-0.8563-0.977-0.0969-0.0474-0.0973 0.093766 5.8-0.8004-0.607-0.0978-0.03895-0.00507 0.09578 5.9-0.76603-0.49-0.089485-0.0367-0.00836 0.09653 6.0-0.737-0.08956-0.08630-0.03370-0.005888 0.09786 6. -0.6894-0.0563-0.08330-0.0340-0.00370 0.099030 6. -0.6407-0.040-0.0808-0.08865-0.006 0.0079 6.3-0.60373-0.09933-0.07745-0.06586 0.000375 0.090 6.4-0.56667-0.096356-0.0747-0.0440 0.0099 0.0334 6.5-0.53083-0.093483-0.07070-0.095 0.00449 0.03340 6.6-0.49635-0.090704-0.069549-0.0084 0.005903 0.0484 6.7-0.4698-0.08807-0.067097-0.08334 0.0076 0.0587 6.8-0.43068-0.085447-0.064730-0.0645 0.00965 0.0605 6.9-0.3996-0.08934-0.06453-0.0464 0.0086 0.06886 7.0-0.36863-0.08053-0.06043-0.0900 0.034 0.07663 7. -0.33958-0.07888-0.05808-0.05 0.03789 0.0844 7. -0.30-0.07590-0.056037-0.009603 0.0594 0.093 7.3-0.8363-0.073745-0.054056-0.00807 0.06553 0.09804
580 PERCENTILES OF SKEW-NORMAL DISTRIBUTIONS Table A. Continued 7.4-0.5676-0.0760-0.05-0.00656 0.07858 0.0468 7.5-0.304-0.069554-0.05040-0.00503 0.093 0.083 7.6-0.0535-0.067574-0.04843-0.003635 0.0036 0.686 7.7-0.806-0.065650-0.04669-0.0076 0.0503 0.49 7.8-0.576-0.063777-0.044988-0.000954 0.063 0.797 7.9-0.347-0.06967-0.043348 0.00039 0.0374 0.333 8.0-0.95-0.0600-0.04749 0.00560 0.04790 0.3809 8. -0.0900-0.058498-0.04000 0.00743 0.05790 0.437 8. -0.06879-0.056837-0.03869 0.003900 0.06798 0.4763 8.3-0.0488-0.0559-0.0374 0.005040 0.07749 0.50 8.4-0.085-0.05366-0.03586 0.0063 0.08656 0.5608 8.5-0.00859-0.0533-0.034446 0.00754 0.09559 0.608 8.6-0.09896-0.05065-0.0335 0.00875 0.03046 0.6405 8.7-0.09705-0.04904-0.0383 0.00979 0.035 0.6774 8.8-0.09590-0.0478-0.030556 0.0049 0.0306 0.76 8.9-0.09353-0.046447-0.0937 0.0070 0.03860 0.7473 9.0-0.098-0.0457-0.0833 0.0977 0.03368 0.7783 9. -0.09059-0.043830-0.0696 0.0845 0.034346 0.800 9. -0.08856-0.04576-0.0584 0.03698 0.035065 0.8397 9.3-0.086938-0.04349-0.04734 0.0453 0.035755 0.8687 9.4-0.085379-0.04058-0.03666 0.05333 0.0364 0.8960 9.5-0.083854-0.038984-0.066 0.068 0.037080 0.9 9.6-0.0838-0.037854-0.0604 0.06868 0.03774 0.9475 9.7-0.08093-0.036748-0.00609 0.07608 0.03838 0.979 9.8-0.079553-0.035683-0.09654 0.0837 0.03897 0.9954 9.9-0.0787-0.03469-0.08706 0.0908 0.039480 0.086 0.0-0.07687-0.033597-0.07784 0.09709 0.040050 0.04 0.5-0.070540-0.0885-0.03538 0.08 0.0464 0.345.0-0.0649-0.04585-0.009749 0.05563 0.044873 0.33 λ α 0.0 0.0 0.05 0.04 0.05 0.0.5-0.05984-0.00784-0.006407 0.07966 0.04689 0.770.0-0.05555-0.0739-0.0034 0.030098 0.04853 0.39.5-0.054-0.0437-0.000708 0.03980 0.05006 0.377 3.0-0.047339-0.055 0.0075 0.033664 0.05355 0.4079 3.5-0.043889-0.009055 0.00398 0.03557 0.0554 0.4365 4.0-0.04074-0.006747 0.005909 0.036500 0.05357 0.4596 4.5-0.0378-0.004656 0.007694 0.037699 0.054480 0.4805 5.0-0.0355-0.0076 0.00935 0.038784 0.055308 0.4958 5.5-0.03685-0.0004 0.00850 0.039764 0.056045 0.5093 6.0-0.030383 0.000586 0.08 0.040657 0.056694 0.500 6.5-0.0868 0.00073 0.03480 0.04444 0.05785 0.58 7.0-0.0630 0.00343 0.04648 0.0477 0.0578 0.5357 7.5-0.0447 0.004693 0.05709 0.0485 0.05893 0.54 8.0-0.0750 0.00585 0.00585 0.06690 0.05875 0.5456
BAGUI & BAGUI 58 Table A. Continued 8.5-0.054 0.00693 0.07599 0.043967 0.059083 0.5477 9.0-0.09658 0.00794 0.08437 0.044467 0.0594 0.5499 9.5-0.0847 0.0089 0.0935 0.04498 0.059745 0.55 0.0-0.06935 0.009766 0.09949 0.045335 0.06008 0.5530 0.5-0.05693 0.00567 0.0064 0.04579 0.06087 0.5547.0-0.0458 0.0346 0.044 0.046078 0.06054 0.5554.5-0.034 0.005 0.0830 0.046386 0.06074 0.556.0-0.0383 0.073 0.0374 0.04668 0.060905 0.5560.5-0.0397 0.03344 0.0886 0.046948 0.06078 0.5566 3.0-0.0047 0.0398 0.03348 0.04700 0.0633 0.5567 3.5-0.00964 0.0448 0.0380 0.04748 0.06370 0.5573 4.0-0.008769 0.04998 0.049 0.047639 0.06500 0.557 4.5-0.007970 0.05477 0.046 0.04783 0.0664 0.5574 5.0-0.00708 0.05950 0.04969 0.048007 0.0676 0.5564 5.5-0.006509 0.0639 0.0535 0.04870 0.06808 0.5573 6.0-0.00583 0.0679 0.05634 0.04833 0.0689 0.5564 6.5-0.00574 0.0776 0.05939 0.048463 0.06967 0.555 7.0-0.004559 0.07544 0.063 0.048588 0.06036 0.5554 7.5-0.003968 0.07898 0.06496 0.048705 0.0609 0.554 8.0-0.0034 0.080 0.06747 0.04883 0.0653 0.5537 8.5-0.00880 0.08530 0.06980 0.04899 0.0605 0.5533 9.0-0.00364 0.088 0.0705 0.0490 0.0645 0.5530 9.5-0.00887 0.090 0.0748 0.049094 0.0686 0.554 30.0-0.004 0.0937 0.0768 0.0498 0.0636 0.5530 3.0-0.000534 0.09867 0.08003 0.04933 0.06388 0.554 3.0-0.00070 0.00300 0.0833 0.049445 0.06445 0.5508 33.0 0.0007 0.0070 0.086 0.049549 0.06486 0.550 34.0 0.00689 0.0065 0.08878 0.049633 0.0653 0.5496 35.0 0.0038 0.0390 0.096 0.0497 0.06557 0.5504 36.0 0.0094 0.0690 0.0933 0.049774 0.06578 0.5490 37.0 0.003455 0.0970 0.095 0.04989 0.0659 0.5496 38.0 0.003950 0.03 0.09697 0.049876 0.0660 0.5493 39.0 0.004439 0.0450 0.0985 0.04998 0.0664 0.5490 40.0 0.004874 0.066 0.09990 0.04995 0.0666 0.5487 45.0 0.00667 0.03457 0.030504 0.050056 0.0664 0.5448 50.0 0.007977 0.03984 0.03089 0.050084 0.0663 0.543
58 PERCENTILES OF SKEW-NORMAL DISTRIBUTIONS Table B. Percentiles of Skew-normal distributions λ α 0.90 0.95 0.96 0.975 0.98 0.99 0.0.9340.656375.76.973.064999.337648 0.0.3056.66439.76995.979043.07866.34487 0.03.309050.67955.7776.986665.080348.35550 0.04.3670.67948.7855.9947.08780.36083 0.05.3480.686977.79677.0053.09586.367557 0.06.3379.694337.79995.00884.0.37450 0.07.33940.70540.80704.05848.0940.3850 0.08.346394.708666.8439.096.6477.38837 0.09.35363.757.833.09747.357.39498 0.0.36077.7577.88044.03643.9964.40439 0.5.39469.75585.86067.068003.60968.43694 0.0.4609.784750.889300.096069.88663.458094 0.5.454635.834.9585.090.3067.480934 0.30.48059.83486.9386.4349.33973.500336 0.35.503898.85553.9584.608.5863.56574 0.40.54704.873564.97578.76768.67009.5300 0.45.543034.889094.9906.90004.7966.540934 0.50.559094.9065.00473.000.9073.54964 0.55.573039.93487.0303.050.98575.55665 0.60.58548.9783.0584.7448.305356.56057 0.65.59544.930570.08706.3336.30856.5665 0.70.60439.936988.034563.80.354.56953 0.75.6804.9438.0399.3836.38608.576 0.80.68075.946476.04889.34565.3086.57340 0.85.63354.949947.04594.36780.3569.574507 0.90.67734.95547.04890.38430.346.57537 0.95.6339.954670.049986.3975.358.575999.0.63440.95643.0549.40476.36339.57675..638840.958590.0537.4585.374.576799..6477.959986.054064.4350.37050.57705.3.6435.96054.054750.443.37984.577070.4.644607.96097.054949.4456.3796.576978.5.645308.9644.059950.4557.37888.576880.6.645648.9666.055057.4478.37804.576944.7.645906.9600.05488.4736.370.576940.8.645948.9658.05507.4478.3785.576950.9.64596.969.054800.463.3799.57696.0.645964.96075.054768.4509.3703.5767..645909.9604.054838.435.3756.576704..64588.960899.05480.409.37690.576740.3.64583.960930.054787.434.37584.576644.4.64579.96085.054695.443.37559.576768.5.64574.960853.054653.46.3776.576606.6.6457.960833.054470.446.36834.57664
BAGUI & BAGUI 583 Table B. Continued.7.645637.96079.054506.43.36950.576564.8.645590.960666.054598.406.3753.576594.9.645570.960694.05449.4087.36953.57658 3.0.645558.960707.05438.4.36884.576578 3..645474.96063.054549.4039.37407.57653 3..645488.96063.05494.43.3677.576504 λ α 0.90 0.95 0.96 0.975 0.98 0.99 3.3.64545.96058.05448.4884.37434.5765 3.4.645466.960594.05465.4045.3660.576445 3.5.645430.96056.054435.4953.373.576440 3.6.645399.960585.05406.435.36654.57657 3.7.645338.960508.054390.499.375.576455 3.8.64534.960490.0548.49.36835.576369 3.9.645379.960450.05497.40.3678.576330 4.0.6453.960373.05438.474.378.576309 4..64535.96044.05454.4898.3665.576407 4..6453.960439.05478.4854.36687.57656 4.3.6456.9604.05433.4768.37075.576 4.4.64509.96044.0547.400.3659.5767 4.5.6458.960400.054053.488.36596.57698 4.6.6456.96099.05446.4636.370.576333 4.7.6456.960338.0540.475.36936.5769 4.8.64508.96035.054063.498.36533.5764 4.9.64563.96080.054060.4765.3663.576 5.0.6454.9604.05405.4575.37085.57606 5..645086.96046.054.4644.36876.57636 5..64565.960308.05399.485.36466.57660 5.3.64535.960306.053968.4907.3639.5766 5.4.645.96053.054074.4737.36709.57684 5.5.645047.9600.05436.4636.36965.57606 5.6.645074.96043.05407.4744.36695.57644 5.7.645095.96086.05394.487.3649.5765 5.8.64507.96033.053909.4780.36384.57643 5.9.645000.96043.05395.4593.366.5768 6.0.64500.96003.054066.445.36907.57698 6..644974.9603.05406.4435.3690.57647 6..645003.9606.053954.4558.36640.576085 6.3.64505.96083.053896.4696.3649.576078 6.4.645045.9607.05387.4734.36339.576060 6.5.645009.9604.05390.468.36459.576040 6.6.64498.960095.053987.4485.36643.57605 6.7.644965.960078.054038.4488.3688.576034 6.8.64495.960053.053990.4437.3684.576046 6.9.644988.960.053900.4568.36596.5766 7.0.644949.96054.053835.4636.3637.576030 7..644964.9607.053805.4699.3686.575996 7..644953.96066.053807.477.3638.576077 7.3.64494.96035.05387.463.3630.575969 7.4.644974.960074.053883.466.3654.57603
584 PERCENTILES OF SKEW-NORMAL DISTRIBUTIONS Table B. Continued 7.5.64497.960047.053996.4583.36695.576053 7.6.644905.96007.053986.4468.3680.575998 7.7.64488.96003.0539.4447.36658.575943 7.8.64499.960073.053898.4500.36506.57593 7.9.64486.960086.053796.4544.36368.5759 8.0.64486.960086.053796.4544.36368.5759 8..644943.96009.053767.463.36.57597 8..644900.96005.05378.4569.3666.57596 8.3.644896.96009.05375.45.36375.575969 8.4.644865.95998.0538.446.36483.575973 λ α 0.90 0.95 0.96 0.975 0.98 0.99 8.5.644835.959963.05384.4366.3664.57603 8.6.644836.959950.05398.47.36747.575987 8.7.64478.959948.05394.436.3684.576077 8.8.644840.959959.053904.4307.36759.576039 8.9.644839.95996.053845.445.36647.575979 9.0.64486.959939.053783.4389.36466.575960 9..644858.959983.05373.446.36388.57598 9..644885.96004.053756.453.365.57598 9.3.64488.9600.053730.4538.364.575987 9.4.644870.96007.053703.4534.3670.57586 9.5.644853.96005.053738.4544.3654.5760 9.6.644877.960038.05370.45.3689.57595 9.7.64480.959976.053759.443.3638.575969 9.8.6448.959973.053800.4396.36444.575875 9.9.644769.959939.05380.4360.36500.57583 0.0.64478.959948.053859.4353.36638.575943 0.5.644804.959943.053758.4444.3639.575953.0.644804.96005.053653.460.3604.575897.5.64478.959947.05368.4464.3686.575794.0.64474.95987.053787.493.36544.575793.5.64476.959876.053757.454.36530.575768 3.0.644757.95999.053645.4376.364.575774 3.5.64474.9599.05358.4459.36090.575783 4.0.64473.9599.05368.4399.3686.57584 4.5.644684.959809.05364.437.36385.575796 5.0.644638.959769.05377.48.36577.57584 5.5.644644.95978.053737.40.36580.575795 6.0.644663.95978.053644.454.36394.575750 6.5.644685.959836.053550.46.36.57574 7.0.644674.959845.05355.4346.36080.575808 7.5.644698.959858.05354.4395.3600.575665 8.0.64470.959889.053565.437.36083.575678 8.5.644645.95987.053607.4333.364.575747 9.0.644663.959800.05369.47.36.57573 9.5.644634.959747.05366.45.36354.575740 0.0.64467.95975.05367.43.3640.57566 0.5.64463.95976.053698.48.36468.57579
BAGUI & BAGUI 585 Table B. Continued.0.64460.9597.053689.46.36337.57555.5.644638.959779.05360.445.3666.575705.0.6446.959777.053609.4303.3648.57563.5.6446.95980.053550.437.3607.575730 3.0.644659.959839.053537.4369.3604.575757 3.5.644647.95984.053499.4389.3594.575648 4.0.64464.959878.053466.445.35935.57569 4.5.644658.95988.053480.4397.35907.5757 5.0.644639.959836.053463.4375.35944.575668 5.5.644635.959838.053506.436.36009.575766 6.5.64467.959805.05358.49.36085.575695 7.0.644580.95978.053498.434.3643.57575 7.5.64457.95977.05353.405.3667.575708 8.0.644597.95973.053560.45.365.57567 8.5.64456.95970.05360.47.3675.575609 λ α 0.90 0.95 0.96 0.975 0.98 0.99 9.0.644596.959695.053639.464.36395.575679 9.5.644554.959687.05365.40.3640.575647 30.0.644567.959746.05364.4064.36440.575569 3.0.64454.959689.053634.4.3644.57568 3.0.64459.959683.05364.4048.3638.575607 33.0.644538.959699.053558.4.3688.57559 34.0.64455.95975.053554.444.3636.575658 35.0.644579.95975.053534.480.3600.5756 36.0.644555.959697.05347.470.360.575533 37.0.644047.959700.053440.49.36077.575647 38.0.644570.959698.053447.48.3606.57569 39.0.644555.95974.05344.488.35948.575530 40.0.644563.959744.05346.43.35948.57564 45.0.64455.959688.05349.403.36003.575680 50.0.64454.959636.05346.4076.3603.575640 Conclusion For any number of runs, the program will calculate the percentiles of skew-normal distributions for a given value of λ. It is advisable that at least one million runs be used to find percentiles. The above table is created using 00 million runs, so these reported results are more precise. One can expect to have better results if the number of runs is very high. From the table it may be noted that for λ =., α = 0.05, Z ( λ ) = Z (.) = 0.69448. α 0.05 However, if λ =., α = 0.05, Z α ( λ) = Z 0.05(.) is not available in the table. Now the value of Z ( λ) = Z (.) can be derived as α 0.05 Z (.) = Z (.) 0.05 0.95 =.95859. Let X ~ SN (3.0,.5,.). Then, X (.) 0.05 can be obtained as X (.) = 3 + (.5) Z 0.05 0.05 (.) = 3 + (.5)(0.69448) =.9588, and X 0.05 (-.) is derived as X (.) = 3 (.5) Z 0.05 0.95 (.) = 3 (.5)(.95859) = 0.065.
586 PERCENTILES OF SKEW-NORMAL DISTRIBUTIONS References Adcock, C., & Shutes, K. (999). Portfolio selection based on the multivariate skewed normal distribution, Financial Modeling, (A Skulimowski edited), Progress and Business Publishers, Krakow, 35-60. Aigner, D. J., & Lovell, C. A. K. (977). Formulation and estimation of stochastic frontier production function model. Econometrics,, -37. Arnold, B. C., & Lin, G. D. (004). Characterization of the skew-normal and generalized Chi distributions. Sankhya, 66(4), 593-606. Azzalini, A. (985). A class of distributions which includes the normal ones. Scandinavian Journal of Statistics,, 7-78. Azzalini, A. (986). Further results on a class of distributions which includes the normal ones. Statistica, 46, 99-08. Azzalini, A., & Dalla valle, A. (996). The multivariate skew-normal distribution. Biometrika, 83(4), 75-76. Azzalini, A., & Capitanio, A. (999). Statistical applications of the multivariate skew normal distribution. Journal of the Royal Statistical Society, B, 6(3), 579-60. Bartolucci, F., De Luca G., & Loperfido, N. (000). A generalization for skewness in the basic stochastic volatility model, Proceedings of the 5 th International Workshop on Statistical Modelling, 40-45. Box, G. E. P. & Muller, M. E. (958). A note on the generation of random normal deviates. Annals of Mathematical Statistics, 9, 60-6. Gupta, A. K., Nguyen, T. T., & Sanqui, J. A. T. (004a). Characterization of the skewnormal distribution. Annals of the Institute of Statistical Mathematics, 56(), 35-360. Gupta, A. K., Gonzalez-Farias, G., & Dominguez-Molina, J. A. (004b). A multivariate skew normal distribution. Journal of Multivariate Analysis, 89, 8-90. O Hagan, A., & Leonard, T. (98). Bayes estimation subject to uncertainty about parameter constraints. Biometrika, 63, 0-03. Roberts, C. (966). A correlation model useful in the study of twins. Journal of the American Statistical Association, 6(36), 84-90. Rubinstein, R. Y. (98). Simulation and the monte carlo method. New York, N.Y.: Wiley & Sons.
BAGUI & BAGUI 587 Appendix Imports System.Math Public Class Form Inherits System.Windows.Forms.Form Dim u, u, z, z, delta, lamda, zscore, num As Single Dim p, p, p3, p4, p5, p6, p7, p8, p9, p0, p, p As Single Dim i, n As Integer Dim file As System.IO.StreamWriter Private Sub Button_Click(ByVal sender As System.Object, ByVal e As System.EventArgs) Handles Button.Click n = Val(TextBox.Text) lamda = Val(TextBox.Text) file = System.IO.File.CreateText("C:\file.txt") Dim zscorearray(n) As Single For i = To n u = Rnd() u = Rnd() z = (Sqrt((-) * Log(u))) * Cos((360) * u) z = (Sqrt((-) * Log(u))) * Sin((360) * u) delta = (lamda / (Sqrt( + (lamda * lamda)))) zscore = (delta * Abs(z) + Sqrt( - (delta * delta)) * z) zscorearray(i) = zscore Next Array.Sort(zScoreArray) p = 0.0 * n p = 0.0 * n p3 = 0.05 * n p4 = 0.04 * n p5 = 0.05 * n p6 = 0. * n p7 = 0.9 * n p8 = 0.95 * n p9 = 0.96 * n p0 = 0.975 * n p = 0.98 * n p = 0.99 * n TextBox3.Text = zscorearray(p) TextBox4.Text = zscorearray(p) TextBox5.Text = zscorearray(p3) TextBox7.Text = zscorearray(p4) TextBox8.Text = zscorearray(p5) TextBox9.Text = zscorearray(p6) TextBox0.Text = zscorearray(p7)
588 PERCENTILES OF SKEW-NORMAL DISTRIBUTIONS TextBox.Text = zscorearray(p8) TextBox.Text = zscorearray(p9) TextBox3.Text = zscorearray(p0) TextBox4.Text = zscorearray(p) TextBox5.Text = zscorearray(p) End Sub file.writeline("number of Runs = " & n & vbcrlf & vbcrlf _ & "Lamda = " & lamda & vbcrlf & vbcrlf _ & "0.0 = " & zscorearray(p) & vbcrlf _ & "0.0 = " & zscorearray(p) & vbcrlf _ & "0.05 = " & zscorearray(p3) & vbcrlf _ & "0.04 = " & zscorearray(p4) & vbcrlf _ & "0.05 = " & zscorearray(p5) & vbcrlf _ & "0. = " & zscorearray(p6) & vbcrlf _ & "0.9 = " & zscorearray(p7) & vbcrlf _ & "0.95 = " & zscorearray(p8) & vbcrlf _ & "0.96 = " & zscorearray(p9) & vbcrlf _ & "0.975 = " & zscorearray(p0) & vbcrlf _ & "0.98 = " & zscorearray(p) & vbcrlf _ & "0.99 = " & zscorearray(p) & vbcrlf & _ vbcrlf _ & "Date and Time is " & System.DateTime.Now) file.close() Private Sub Form_Load(ByVal sender As System.Object, ByVal e As System.EventArgs) Handles MyBase.Load Randomize() 'Calling the random number generator End Sub End Class