Convergence in Distribution Central Limit Theorem

Total Page:16

File Type:pdf, Size:1020Kb

Convergence in Distribution Central Limit Theorem Convergence in Distribution Central Limit Theorem Statistics 110 Summer 2006 Copyright °c 2006 by Mark E. Irwin Convergence in Distribution Theorem. Let X » Bin(n; p) and let ¸ = np, Then µ ¶ n e¡¸¸x lim P [X = x] = lim px(1 ¡ p)n¡x = n!1 n!1 x x! So when n gets large, we can approximate binomial probabilities with Poisson probabilities. Proof. µ ¶ µ ¶ µ ¶ µ ¶ n n ¸ x ¸ n¡x lim px(1 ¡ p)n¡x = lim 1 ¡ n!1 x n!1 x n n µ ¶ µ ¶ µ ¶ n! 1 ¸ ¡x ¸ n = ¸x 1 ¡ 1 ¡ x!(n ¡ x)! nx n n Convergence in Distribution 1 µ ¶ µ ¶ µ ¶ n! 1 ¸ ¡x ¸ n = ¸x 1 ¡ 1 ¡ x!(n ¡ x)! nx n n µ ¶ ¸x n! 1 ¸ n = lim 1 ¡ x! n!1 (n ¡ x)!(n ¡ ¸)x n | {z } | {z } !1 !e¡¸ e¡¸¸x = x! 2 Note that approximation works better when n is large and p is small as can been seen in the following plot. If p is relatively large, a di®erent approximation should be used. This is coming later. (Note in the plot, bars correspond to the true binomial probabilities and the red circles correspond to the Poisson approximation.) Convergence in Distribution 2 lambda = 1 n = 10 p = 0.1 lambda = 1 n = 50 p = 0.02 lambda = 1 n = 200 p = 0.005 p(x) p(x) p(x) 0.0 0.1 0.2 0.3 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 x x x lambda = 5 n = 10 p = 0.5 lambda = 5 n = 50 p = 0.1 lambda = 5 n = 200 p = 0.025 p(x) p(x) p(x) 0.00 0.05 0.10 0.15 0.20 0.00 0.05 0.10 0.15 0.00 0.05 0.10 0.15 0 1 2 3 4 5 6 7 8 9 0 2 4 6 8 10 12 0 2 4 6 8 10 12 x x x Convergence in Distribution 3 Example: Let Y1;Y2;::: be iid Exp(1). Then Xn = Y1 + Y2 + ::: + Yn » Gamma(n; 1) which has p E[Xn] = n; Var(Xn) = n; SD(Xn) = n Xpn¡n Thus Zn = n has mean = 0 and variance = 1. Lets compare its distribution to Z » N(0; 1). i.e. Is P [¡1 · Zn · 2] ¼ P [¡1 · Z · 2]? Let X ¡ n p Z = np ; X = n + nZ n n n n p p fZn(z) = fXn(n + nz) £ n Convergence in Distribution 4 Z b P [a · Zn · b] = fZn(z)dz a Z b p p = nfXn(n + nz)dz a Z p b p (n + nz)n¡1 p = n e¡(n+ nz)dz a (n ¡ 1)! p To go further we need Stirling's Formula: n! ¼ nne¡n 2¼n. So p p p p p n f (n + nz) n = e¡n¡z n(n + z n)n¡1 Xn (n ¡ 1)! p p p e¡n¡z n(n + z n)n¡1 n ¼ p (n ¡ 1)n¡1e¡n+1 2¼n µ ¶ 1 p z n ¼ p e¡z n 1 + p n 2¼ | {z } gn(z) Convergence in Distribution 5 µ ¶ p z log(g (z)) = ¡z n + n log 1 + p n n · ¸ µ ¶ p z 1z2 1 z3 1 1 = ¡z n + n p ¡ + ¡ ::: ¼ ¡ z2 + O p n 2 n 3n3=2 2 n so p p 1 ¡z2=2 fX (n + z n) n ¼ p e n 2¼ Thus Z b 1 ¡z2=2 P [a · Zn · b] ! p e dz = P [a · Z · b] a 2¼ So as n increases, the distribution of Zn gets closer and closer to a N(0; 1). Convergence in Distribution 6 p Another way of thinking of this, is that the distribution of Xn = n + Zn n approaches that of a N(n; n). n = 2 n = 5 n = 10 f(x) f(x) f(x) 0.0 0.1 0.2 0.3 0.00 0.05 0.10 0.15 0.20 0.00 0.04 0.08 0.12 −2 0 2 4 6 −2 0 2 4 6 8 10 12 0 5 10 15 20 x x x n = 20 n = 50 n = 100 f(x) f(x) f(x) 0.00 0.02 0.04 0.06 0.08 0.00 0.01 0.02 0.03 0.04 0.05 0.00 0.01 0.02 0.03 0.04 5 10 15 20 25 30 35 30 40 50 60 70 70 80 90 100 110 120 130 x x x Convergence in Distribution 7 De¯nition. Let X1;X2;::: be a sequence of RVs with cumulative distribution functions F1;F2;::: and let X be a RV with distribution F . We say Xn Converges in Distribution to X if lim Fn(x) = F (x) n!1 D at every point at which F is continuous. Xn ¡! X An equivalent statement to this is that for all a and b where F is continuous P [a · Xn · b] ! P [a · X · b] Note that if Xn and X are discrete distributions, this condition reduces to P [Xn = xi] ! P [X = xi] for all support points xi. Convergence in Distribution 8 Note that an equivalent de¯nition of convergence in distribution is that D Xn ¡! X if E[g(Xn)] ! E[g(X)] for all bounded, continuous functions g(¢). This statement of convergence in distribution is needed to help prove the following theorem Theorem. [Continuity Theorem] Let Xn be a sequence of random variables with cumulative distribution functions Fn(x) and corresponding moment generating functions Mn(t). Let X be a random variable with cumulative distribution function F (x) and moment generating function M(t). If Mn(t) ! M(t) for all t in an open interval containing zero, then D Fn(x) ! F (x) at all continuity points of F . That is Xn ¡! X. Thus the previous two examples (Binomial/Poisson and Gamma/Normal) could be proved this way. Convergence in Distribution 9 For the Gamma/Normal example µ ¶ à !n p p t ¡t n 1 ¡t n MZ (t) = MX p e = e n n pt n 1 ¡ n Similarly to the earlier proof, its easier to work with log MZn(t) µ ¶ p t log M (t) = ¡t n ¡ n log 1 ¡ p Zn n · ¸ p t 1t2 1 t3 = ¡t n ¡ n ¡p ¡ ¡ ¡ ::: n 2 n 3n3=2 µ ¶ 1 1 = t2 + O p 2 n Thus t2=2 MZn(t) ! e which is the MGF for a standard normal. Convergence in Distribution 10 Central Limit Theorem Theorem. [Central Limit Theorem (CLT)] Let X1;X2;X3;::: be a sequence of independent RVs having mean ¹ and variance σ2 and a common distribution function F (x) and moment generating function M(t) de¯ned in a neighbourhood of zero. Let Xn Sn = Xn i=1 Then · ¸ S ¡ n¹ lim P n p · x = ©(x) n!1 σ n That is S ¡ n¹ n p ¡!D N(0; 1) σ n Central Limit Theorem 11 Proof. Without a loss of generality, we can assume that ¹ = 0. So let Spn Zn = σ n. Since Sn is the sum of n iid RVs, µ µ ¶¶ t n M (t) = (M(t))n ; M (t) = M p Sn Zn σ n Taking a Taylor series expansion of M(t) around 0 gives 1 1 M(t) = M(0) + M 0(0)t + M 00(0)t2 + ² = 1 + σ2t2 + O(t3) 2 t 2 since M(0) = 1;M 0(0) = ¹ = 0;M 00(0) = σ2. So µ ¶ µ ¶ õ ¶ ! t 1 t 2 t 3 M p = 1 + σ2 p + O p σ n 2 σ n σ n µ ¶ t2 1 = 1 + + O 2n n3=2 Central Limit Theorem 12 This gives µ 2 µ ¶¶n t 1 2 M (t) = 1 + + O ! et =2 Zn 2n n3=2 2 Note that the requirement of a MGF is not needed for the theorem to hold. 2 In fact, all that is needed is that Var(Xi) = σ < 1. A standard proof of this more general theorem uses the characteristic function (which is de¯ned for any distribution) Z 1 Á(t) = eitxf(x)dx = M(it) ¡1 p instead of the moment generating function M(t), where i = ¡1. Thus the CLT holds for distributions such as the log normal, even though it doesn't have a MGF. Central Limit Theorem 13 Also, the CLT is often presented in the following equivalent form X¹ ¡ ¹ p X¹ ¡ ¹ Z = n p = n n ¡!D N(0; 1) n σ= n σ To see this is the same, just multiply the numerator and denominator by n in the ¯rst form to get the statement about Sn. The common way that this is used is that µ 2¶ approx: ¡ ¢ approx: σ S » N n¹; nσ2 or X¹ » N ¹; n n n Central Limit Theorem 14 Example: Insurance claims Suppose that an insurance company has 10,000 policy holders. The expected yearly claim per policyholder is $240 with a standard deviation of $800. What is the approximate probability that the total yearly claims S10;000 > $2.6 Million E[S ] = 10; 000 £ 240 = 2; 400; 000 10;000 p SD(S10;000) = 10; 000 £ 800 = 80; 000 P [S10;000 > 2; 600; 000] · ¸ S ¡ 2; 400; 000 2; 600; 000 ¡ 2; 400; 000 = P 10;000 > 80; 000 80; 000 ¼ P [Z > 2:5] = 0:0062 Note that this probability statement does not use anything about the distribution of the original policy claims except their mean and standard deviation.
Recommended publications
  • Use of Proc Iml to Calculate L-Moments for the Univariate Distributional Shape Parameters Skewness and Kurtosis
    Statistics 573 USE OF PROC IML TO CALCULATE L-MOMENTS FOR THE UNIVARIATE DISTRIBUTIONAL SHAPE PARAMETERS SKEWNESS AND KURTOSIS Michael A. Walega Berlex Laboratories, Wayne, New Jersey Introduction Exploratory data analysis statistics, such as those Gaussian. Bickel (1988) and Van Oer Laan and generated by the sp,ge procedure PROC Verdooren (1987) discuss the concept of robustness UNIVARIATE (1990), are useful tools to characterize and how it pertains to the assumption of normality. the underlying distribution of data prior to more rigorous statistical analyses. Assessment of the As discussed by Glass et al. (1972), incorrect distributional shape of data is usually accomplished conclusions may be reached when the normality by careful examination of the values of the third and assumption is not valid, especially when one-tail tests fourth central moments, skewness and kurtosis. are employed or the sample size or significance level However, when the sample size is small or the are very small. Hopkins and Weeks (1990) also underlying distribution is non-normal, the information discuss the effects of highly non-normal data on obtained from the sample skewness and kurtosis can hypothesis testing of variances. Thus, it is apparent be misleading. that examination of the skewness (departure from symmetry) and kurtosis (deviation from a normal One alternative to the central moment shape statistics curve) is an important component of exploratory data is the use of linear combinations of order statistics (L­ analyses. moments) to examine the distributional shape characteristics of data. L-moments have several Various methods to estimate skewness and kurtosis theoretical advantages over the central moment have been proposed (MacGillivray and Salanela, shape statistics: Characterization of a wider range of 1988).
    [Show full text]
  • A Skew Extension of the T-Distribution, with Applications
    J. R. Statist. Soc. B (2003) 65, Part 1, pp. 159–174 A skew extension of the t-distribution, with applications M. C. Jones The Open University, Milton Keynes, UK and M. J. Faddy University of Birmingham, UK [Received March 2000. Final revision July 2002] Summary. A tractable skew t-distribution on the real line is proposed.This includes as a special case the symmetric t-distribution, and otherwise provides skew extensions thereof.The distribu- tion is potentially useful both for modelling data and in robustness studies. Properties of the new distribution are presented. Likelihood inference for the parameters of this skew t-distribution is developed. Application is made to two data modelling examples. Keywords: Beta distribution; Likelihood inference; Robustness; Skewness; Student’s t-distribution 1. Introduction Student’s t-distribution occurs frequently in statistics. Its usual derivation and use is as the sam- pling distribution of certain test statistics under normality, but increasingly the t-distribution is being used in both frequentist and Bayesian statistics as a heavy-tailed alternative to the nor- mal distribution when robustness to possible outliers is a concern. See Lange et al. (1989) and Gelman et al. (1995) and references therein. It will often be useful to consider a further alternative to the normal or t-distribution which is both heavy tailed and skew. To this end, we propose a family of distributions which includes the symmetric t-distributions as special cases, and also includes extensions of the t-distribution, still taking values on the whole real line, with non-zero skewness. Let a>0 and b>0be parameters.
    [Show full text]
  • A Family of Skew-Normal Distributions for Modeling Proportions and Rates with Zeros/Ones Excess
    S S symmetry Article A Family of Skew-Normal Distributions for Modeling Proportions and Rates with Zeros/Ones Excess Guillermo Martínez-Flórez 1, Víctor Leiva 2,* , Emilio Gómez-Déniz 3 and Carolina Marchant 4 1 Departamento de Matemáticas y Estadística, Facultad de Ciencias Básicas, Universidad de Córdoba, Montería 14014, Colombia; [email protected] 2 Escuela de Ingeniería Industrial, Pontificia Universidad Católica de Valparaíso, 2362807 Valparaíso, Chile 3 Facultad de Economía, Empresa y Turismo, Universidad de Las Palmas de Gran Canaria and TIDES Institute, 35001 Canarias, Spain; [email protected] 4 Facultad de Ciencias Básicas, Universidad Católica del Maule, 3466706 Talca, Chile; [email protected] * Correspondence: [email protected] or [email protected] Received: 30 June 2020; Accepted: 19 August 2020; Published: 1 September 2020 Abstract: In this paper, we consider skew-normal distributions for constructing new a distribution which allows us to model proportions and rates with zero/one inflation as an alternative to the inflated beta distributions. The new distribution is a mixture between a Bernoulli distribution for explaining the zero/one excess and a censored skew-normal distribution for the continuous variable. The maximum likelihood method is used for parameter estimation. Observed and expected Fisher information matrices are derived to conduct likelihood-based inference in this new type skew-normal distribution. Given the flexibility of the new distributions, we are able to show, in real data scenarios, the good performance of our proposal. Keywords: beta distribution; centered skew-normal distribution; maximum-likelihood methods; Monte Carlo simulations; proportions; R software; rates; zero/one inflated data 1.
    [Show full text]
  • Approximating the Distribution of the Product of Two Normally Distributed Random Variables
    S S symmetry Article Approximating the Distribution of the Product of Two Normally Distributed Random Variables Antonio Seijas-Macías 1,2 , Amílcar Oliveira 2,3 , Teresa A. Oliveira 2,3 and Víctor Leiva 4,* 1 Departamento de Economía, Universidade da Coruña, 15071 A Coruña, Spain; [email protected] 2 CEAUL, Faculdade de Ciências, Universidade de Lisboa, 1649-014 Lisboa, Portugal; [email protected] (A.O.); [email protected] (T.A.O.) 3 Departamento de Ciências e Tecnologia, Universidade Aberta, 1269-001 Lisboa, Portugal 4 Escuela de Ingeniería Industrial, Pontificia Universidad Católica de Valparaíso, Valparaíso 2362807, Chile * Correspondence: [email protected] or [email protected] Received: 21 June 2020; Accepted: 18 July 2020; Published: 22 July 2020 Abstract: The distribution of the product of two normally distributed random variables has been an open problem from the early years in the XXth century. First approaches tried to determinate the mathematical and statistical properties of the distribution of such a product using different types of functions. Recently, an improvement in computational techniques has performed new approaches for calculating related integrals by using numerical integration. Another approach is to adopt any other distribution to approximate the probability density function of this product. The skew-normal distribution is a generalization of the normal distribution which considers skewness making it flexible. In this work, we approximate the distribution of the product of two normally distributed random variables using a type of skew-normal distribution. The influence of the parameters of the two normal distributions on the approximation is explored. When one of the normally distributed variables has an inverse coefficient of variation greater than one, our approximation performs better than when both normally distributed variables have inverse coefficients of variation less than one.
    [Show full text]
  • Portfolio Allocation with Skewness Risk: a Practical Guide∗
    Portfolio Allocation with Skewness Risk: A Practical Guide∗ Edmond Lezmi Hassan Malongo Quantitative Research Quantitative Research Amundi Asset Management, Paris Amundi Asset Management, Paris [email protected] [email protected] Thierry Roncalli Rapha¨elSobotka Quantitative Research Multi-Asset Management Amundi Asset Management, Paris Amundi Asset Management, Paris [email protected] [email protected] February 2019 (First version: June 2018) Abstract In this article, we show how to take into account skewness risk in portfolio allocation. Until recently, this issue has been seen as a purely statistical problem, since skewness corresponds to the third statistical moment of a probability distribution. However, in finance, the concept of skewness is more related to extreme events that produce portfolio losses. More precisely, the skewness measures the outcome resulting from bad times and adverse scenarios in financial markets. Based on this interpretation of the skewness risk, we focus on two approaches that are closely connected. The first one is based on the Gaussian mixture model with two regimes: a `normal' regime and a `turbulent' regime. The second approach directly incorporates a stress scenario using jump-diffusion modeling. This second approach can be seen as a special case of the first approach. However, it has the advantage of being clearer and more in line with the experience of professionals in financial markets: skewness is due to negative jumps in asset prices. After presenting the mathematical framework, we analyze an investment portfolio that mixes risk premia, more specifically risk parity, momentum and carry strategies. We show that traditional portfolio management based on the volatility risk measure is biased and corresponds to a short-sighted approach to bad times.
    [Show full text]
  • Skewness and Kurtosis 2
    A distribution is said to be 'skewed' when the mean and the median fall at different points in the distribution, and the balance (or centre of gravity) is shifted to one side or the other-to left or right. Measures of skewness tell us the direction and the extent of Skewness. In symmetrical distribution the mean, median and mode are identical. The more the mean moves away from the mode, the larger the asymmetry or skewness A frequency distribution is said to be symmetrical if the frequencies are equally distributed on both the sides of central value. A symmetrical distribution may be either bell – shaped or U shaped. 1- Bell – shaped or unimodel Symmetrical Distribution A symmetrical distribution is bell – shaped if the frequencies are first steadily rise and then steadily fall. There is only one mode and the values of mean, median and mode are equal. Mean = Median = Mode A frequency distribution is said to be skewed if the frequencies are not equally distributed on both the sides of the central value. A skewed distribution may be • Positively Skewed • Negatively Skewed • ‘L’ shaped positively skewed • ‘J’ shaped negatively skewed Positively skewed Mean ˃ Median ˃ Mode Negatively skewed Mean ˂ Median ˂ Mode ‘L’ Shaped Positively skewed Mean ˂ Mode Mean ˂ Median ‘J’ Shaped Negatively Skewed Mean ˃ Mode Mean ˃ Median In order to ascertain whether a distribution is skewed or not the following tests may be applied. Skewness is present if: The values of mean, median and mode do not coincide. When the data are plotted on a graph they do not give the normal bell shaped form i.e.
    [Show full text]
  • Long-Term Skewness and Systemic Risk
    Journal of Financial Econometrics, 2011, Vol. 9, No. 3, 437–468 Long-Term Skewness and Systemic Risk ROBERT F. ENGLE Stern School of Business at New York University ABSTRACT Financial risk management has generally focused on short-term risks rather than long-term risks, and arguably this was an important component of the recent financial crisis. Econometric approaches to measuring long-term risk are developed in order to estimate the term structure of value at risk and expected shortfall. Long-term negative Downloaded from skewness increases the downside risk and is a consequence of asym- metric volatility models. A test is developed for long-term skewness. In a Merton style structural default model, bankruptcies are accompanied jfec.oxfordjournals.org by substantial drops in equity prices. Thus, skewness in a market factor implies high defaults and default correlations even far in the future cor- roborating the systemic importance of long-term skewness. Investors concerned about long-term risks may hedge exposure as in the In- at New York University on September 20, 2011 tertemporal Capital Asset Pricing Model (ICAPM). As a consequence, the aggregate wealth portfolio should have asymmetric volatility and hedge portfolios should have reversed asymmetric volatility. Using es- timates from VLAB, reversed asymmetric volatility is found for many possible hedge portfolios such as volatility products, long- and short- term treasuries, some exchange rates, and gold. ( JEL: G01) KEYWORDS: ARCH, GARCH, Hedge portfolios, long-term risk, ICAPM, skewness, systemic risk The financial crisis that engulfed the global economy in 2008 and 2009 has been described and diagnosed by many. A thoughtful discussion of the causes and remedies can be found in Acharya and Richardson (2009), Acharya et al.
    [Show full text]
  • Mean-Variance-Skewness Portfolio Performance Gauging: a General Shortage Function and Dual Approach
    EDHEC RISK AND ASSET MANAGEMENT RESEARCH CENTRE 393-400 promenade des Anglais 06202 Nice Cedex 3 Tel.: +33 (0)4 93 18 32 53 E-mail: [email protected] Web: www.edhec-risk.com Mean-Variance-Skewness Portfolio Performance Gauging: A General Shortage Function and Dual Approach September 2005 Walter Briec Maître de conférences, University of Perpignan Kristiaan Kersten Chargé de recherche, CNRS-LABORES, IESEG Octave Jokung Professor, EDHEC Business School Abstract This paper proposes a nonparametric efficiency measurement approach for the static portfolio selection problem in mean-variance-skewness space. A shortage function is defined that looks for possible increases in return and skewness and decreases in variance. Global optimality is guaranteed for the resulting optimal portfolios. We also establish a link to a proper indirect mean-variance-skewness utility function. For computational reasons, the optimal portfolios resulting from this dual approach are only locally optimal. This framework makes it possible to differentiate between portfolio efficiency and allocative efficiency, and a convexity efficiency component related to the difference between the primal, non-convex approach and the dual, convex approach. Furthermore, in principle, information can be retrieved about the revealed risk aversion and prudence of investors. An empirical section on a small sample of assets serves as an illustration. JEL: G11 Keywords: shortage function, efficient frontier, mean-variance-skewness portfolios, risk aversion, prudence. EDHEC is one of the top five business schools in France. Its reputation is built on the high quality of its faculty (110 professors and researchers from France and abroad) and the privileged relationship with professionals that the school has cultivated since its establishment in 1906.
    [Show full text]
  • Statistical Notes for Clinical Researchers: Assessing Normal Distribution (2) Using Skewness and Kurtosis
    Open lecture on statistics ISSN 2234-7658 (print) / ISSN 2234-7666 (online) http://dx.doi.org/10.5395/rde.2013.38.1.52 Statistical notes for clinical researchers: assessing normal distribution (2) using skewness and kurtosis Hae-Young Kim As discussed in the previous statistical notes, although many statistical methods have been proposed to test normality of data in various ways, there is no current Department of Dental Laboratory gold standard method. The eyeball test may be useful for medium to large sized Science & Engineering, Korea (e.g., n > 50) samples, however may not useful for small samples. The formal University College of Health normality tests including Shapiro-Wilk test and Kolmogorov-Smirnov test may be Science, Seoul, Korea used from small to medium sized samples (e.g., n < 300), but may be unreliable for large samples. Moreover we may be confused because ‘eyeball test’ and ‘formal normality test’ may show incompatible results for the same data. To resolve the problem, another method of assessing normality using skewness and kurtosis of the distribution may be used, which may be relatively correct in both small samples and large samples. 1) Skewness and kurtosis Skewness is a measure of the asymmetry and kurtosis is a measure of ’peakedness’ of a distribution. Most statistical packages give you values of skewness and kurtosis as well as their standard errors. In SPSS you can find information needed under the following menu: Analysis – Descriptive Statistics – Explore *Correspondence to Hae-Young Kim, DDS, PhD. Associate Professor, Department of Dental Laboratory Science & Engineering, Korea Skewness is a measure of the asymmetry of the distribution of a variable.
    [Show full text]
  • Understanding the Forecast Statistics and Four Moments ROV Technical Papers Series: Volume 3
    Understanding the Forecast Statistics and Four Moments www.realoptionsvaluation.com ROV Technical Papers Series: Volume 3 The Four Moments In This Issue Most distributions can be defined up to four moments. The first moment describes its location or central tendency (expected values); the second moment describes its width or 1. Learn the basics of spread (risks and uncertainties); the third moment, its directional skew (most probable statistical moments events); and the fourth moment, its peakedness or thickness in the tails (catastrophic extreme 2. Understand the first four tail events). All four moments should be calculated in practice and interpreted to provide a moments of a distribution more comprehensive view of the project under analysis. Risk Simulator provides the results 3. Run correlated simulations of all four moments in its Statistics view in the forecast charts. in Risk Simulator 4. Review the correlations Measuring the Center of the Distribution: The First Moment copulas and random The first moment of a distribution measures the expected value or expected rate of return on number generators in Risk a particular project. It measures the location of the project’s scenarios and possible outcomes Simulator on average. The common statistics for the first moment include the mean (average), median (center of a distribution), and mode (most commonly occurring value). Figure 1 illustrates the first moment––where, in this case, the first moment of this distribution is measured by the mean (), or average, value. “What are the effects of correlations on risk and how 1 1 = 2 2 do you diversify a portfolio with Skew = 0 correlations?” KurtosisXS = 0 1 1 ≠ 2 2 Figure 1.
    [Show full text]
  • The Skewness of the Stock Market at Long Horizons
    The Skewness of the Stock Market at Long Horizons Anthony Neuberger and Richard Payne Cass Business School, City, University of London January 2018 Abstract Moments of long-horizon returns are important for asset pricing but are hard to measure. Proxies for these moments are often used but none is wholly satisfactory. We show analytically that short-horizon (i.e. daily) returns can be used to make more much precise estimates of long-horizon (e.g. annual) moments without making strong assumptions about the data generating process. Skewness comprises two components: the skewness of short-horizon returns, and a leverage effect, which is the covariance between contemporaneous variance and lagged returns. We provide similar results for kurtosis. Applying the technology to US stock index returns, we show that skew is large and negative and does not significantly attenuate with horizon as one goes from monthly to multi-year horizons. - 1 - INTRODUCTION This paper makes two contributions: methodological and empirical. The methodological contribution is to show how short horizon returns can be used to estimate the higher moments of long horizon returns while making only weak assumptions about the data generating process. The empirical contribution is to show that long horizon (multi-year) US equity market returns are highly negatively skewed. The skew coefficient, at around -1.5, is economically significant. We also show that this skew at long horizons is entirely attributable to the leverage effect – the negative correlation between returns and future volatility. There is good reason to believe that higher moments of returns – not just second moments – are important for asset pricing.
    [Show full text]
  • Measures of Multivariate Skewness and Kurtosis in High-Dimensional Framework
    Measures of multivariate skewness and kurtosis in high-dimensional framework Takuma SUMIKAWA∗ Kazuyuki KOIZUMIy Takashi SEOz Abstract In this paper, we propose new definitions for multivariate skewness and kurtosis when the covariance matrix has a block diagonal structure. These measures are based on the ones of Mardia (1970). We give the expectations and the variances for new multivariate sample mea- sures of skewness and kurtosis. Further, we derive asymptotic distributions of statistics by new measures under multivariate normality. To evaluate accuracy of these statistics, numer- ical results are given by the Monte Carlo simulation. We consider the problem estimating for covariance structure. Pavlenko, Bj¨orkstr¨omand Tillander (2012) consider a method which approximate the inverse covariance matrix to block diagonal structure via gLasso. In this paper, we propose an improvement of Pavlenko, Bj¨orkstr¨omand Tillander (2012)'s method by using AIC. Finally, numerical results are shown in order to investigate probability that the new method select true model for covariance structure. Key Words and Phrases: Multivariate Skewness; Multivariate Kurtosis; Normality test; Co- variance structure approximation; Block diagonal matrix; AIC. 1 Introduction In multivariate statical analysis, normality for sample is assumed in many cases. Hence, assessing for multivariate normality is an important problem. The graphical method and the statistical hypothesis testing are considered to this problem by many authors (see, e.g. Henze (2002); Thode (2002)). From aspects of calculation cost and simplicity, we focus on the testing theory based on multivariate skewness and kurtosis. There are various definitions of multivariate skewness and kurtosis (Mardia (1970), Malkovich and Afifi (1973), Srivastava (1984) and so on.).
    [Show full text]