MOMENTS, SKEWNESS & KURTOSIS ‘Moments: In statistics, moment, ane certain constant values in a given distribution, The moments help us to ascertain the nature an! for the underlying distribution. Moments of a distribution may be calculated from arithmetic mean of the distribution or from any arbitrary chosen value including vo (origin). When the moments are computed from the arithmetic mean Of the distribution, we cll hin oments about mean or central moments. When they are computed from am arbitrary value, we call thom raw moments. When they are computed from zero, they are called moment about origin. Moments about origin areal called raw moments 1x0, ne ben observativins ofa variable, then the rth raw moment is defined by, + where A is ony arbitrary value The rth central moment is le! inv! by Le. -2 Tht, Xeon Occur with frequencies ff... respectively then the rth raw moment is Sh -ar A a) =, where, n= ¥° frand A is any arbitrary value. ‘And the rth central moment is Sin sO Dae ~A) Drew sand so on, My My x, Compute the firs three contra moments fr the following frequency distribution [x [2]3]4[5 [6 thst pt Solution, We prepare the follow ng table for computing the moments LGD | H.-F | LG, -3 ale Here, ¥‘MOMENTS, SKEWNESS & KURTOSIS ie Do, -7 Thus, jy = 4 Relationship between raw moments and central moments, Let Xbe a ranciom variable assuming values Xi, X2.-...% with mean ot be any arbitrary value, then Buy. La-m Sea 38 __a- aes = Hs Baa #3? U8)" = Hh Soa + 2a) simulans aa = A 4s +635 (06)? Ho) Sheppard's C for tis generally assumed tho! the froquencies in a group are concentrated at the mid point of the class interval. Ths is surcly a» approximation. W.F Sheppard eorved that if) the frequency distribution is continuous and b te -yueney tapers off to zero in both “Trecions ofthe fequeney dstebution, then the corection fr differ inoments due to grouping at the sri point ofthe class interval are done bythe ollsng formula, know as Sheppard's correction, Corrected) = ty (Correcte 7 1 (Corrected) = Hy EC n+ LC ywhere, Cis the length of class interval tCorvected) = My 3C 3a 8 skewness, Skowness mean ack of symmetry that s departure for symmetry ofa distribution. A listribution i sai to be skewed if i) Mean, Median and Mode give different values, fi) rand Quare not equidistant from the media. {iy The carve dawn with heel ofthe given data snot yma ut tured nas to one se than the other. , by | i walee NOMENTS, SKEWNESS & KURTOSIS esr 2) Sommetieal Dang skswnes aa frequency distribution distribution is Known he type of the distribution, Mean = Media = Mode, o b) Positively Skewed Distibution, In this distribution, the long tail to hv “ right indicates the Presence of extreme values ot she positive end of the distribution. This type ©! a distribution is known as positively skewes listribution. These distributions occur with, (or example, family size, female age at marriage, waxes of employees ete. For this type of distribution, Mean > Median > Mode. ish SheredGoive Sherman) ©) Negatively Skewed Distribution, Ina negatively Mote skewed distribution, the mean is pulled in a —— negative direction. Daily maximum temperature for a month in winter will result in such a negatively skewed curve. lof this type of distribution, Mean < Median < Mode, et Shewed epnre Shewaes) Measures of Skewness. SD. 2 Pearson's Coefficient of skewness = If Mean > Mode, the skew is positive, If Mean < Mode, the skew is neyative, If Mean = Mode, the skew is “ero, in which case the distribution is symmetrical. For a moderately skewed distribution, Mean - Mode = 3(Mean - Median) fean~ Median “Therefore, Pearson's Coefficient of skewness = Mean Median) Quartile Coefficient of skew nes “a {A relative measure of skewness denoted by is defined as A, = 5 Instead of f,, Karl Pearson suggested 7; to be used as a measure of skewness, ai Where, 1 = Vi = 7 Obviously, for a symmetrical distribution, 7MoMENTS, SKEWNESS & KURTOSTS w a normal distribution is called wn as leptokurtic. ithe curve It is called Kurtoss. The degre of peak ness o flatness ofa distribution relative kurtosis. A curve having relatively higher peak than the no! is more flat-topped than the normal curve, itis called esokurtic, whichis nether too peaked nor too lat=topPed mal curve, is kno platykurtc. normal curve itsel Lepokuane Meso — Measures of Kuti, The mst important mesure of kurtosis based on 2n!and 4% moments is Br, ipand pare respectively the second en forth moments about the mean. defined as f, i ‘This measure isa pure number and is always positive. If 23 the distribution is leptokurtic. If i, <3 the distribution is platykurtic If B,=3 the distribution is mesokurtic. Ex Fora distribution, the four central moment and j1, =[Link] f, and f,and hence comment on the nature of the distribution. is were found to be as follows: 44, =0, 2 _ (0.7) 1875 so f= He = OD no031 and f= He Oar ann as A es) asd onthe values of fan fi, we conclude thatthe distribution slightly positively skewed and ante ‘Theorem. For any set of observations, x1, X2.-1.r%n prove that () Bs = fi, +1 and (ii) By 21 Xe. a Proof. () We know, 11 Le Consider the folowing expression, falx, 3) +(x, - 3) +e)" where ab carearlary constant. Ifa, bande are assumed tobe el hen tneabove ls alays positive. ‘Thus performing the squares summing and dividing by nal through ade, PYG, -v (x, — ¥)" x, - F x, = ¥) Deen 2 PEO ag gp Dy rae AAD , WH 2 Sal, +b, +e +2abje, + 2acyy +020 Choosing ant, £2. and == ny, theaboveexpresion becomes,NOMBNTS, SKEWNESS & KURTOSIS Dividing both sides by yy MH izo wom > fy-B,-120 2B, zA + (i) Let z, = 0, 0) summing both sides of (1) an lviding throughout by n L == @) ‘and 2 from (2) and (3) maa 20 2130 # 2A Substituting the values of