Probability Transforms with Elliptical Generators

Size: px
Start display at page:

Download "Probability Transforms with Elliptical Generators"

Transcription

1 Probability Transforms wit Elliptical Generators Emiliano A. Valdez Scool of Actuarial Studies Faculty of Commerce & Economics University of New Sout Wales Sydney, AUSTRALIA Tel Fax First Draft: 9 Marc 2005 Tis Draft: 1 May 2005 Abstract Tis paper introduces te notion of elliptical transformations for possible applications in constructing insurance premium principles. It is a common notion in actuarial science to transform or sometimes said, to distort te real probability measure of a risk X and ten calculate premiums based on te expectation of te transformed distribution. Elliptical transformations are a way to distort te probability distribution of a risk and it is based on te relative ratio of a density generator of a member of te class of elliptical distributions to te density generator of te Normal distribution wic is also well-known to belong to te class of elliptical distributions. Te class of elliptical distribution models consists of distributions considered symmetric and provides a generalization of te class of Normal loss distribution models. We examine te premium principle implied by tis elliptical transformation. We find tat te elliptical transforms lead us to, as special cases, te Wang premium principle, te Wang Student-t distortion principle, as well as te Esscer premium principle. In several cases, te resulting premium principle as a replicating feature of te standard deviation premium principle. A numerical example as been elaborated to illustrate te practical implementation of tis elliptical transformation. Keywords: probability transform, distortion, risk measures, premium principles, elliptical distributions, elliptical transforms. Te autor wises to acknowledge financial support from te Australian Researc Council Discovery Grant DP and te UNSW Actuarial Foundation of Te Institute of Actuaries of Australia. Te autor also extends appreciation to Prof. Dr. Jan Daene of Katolieke Universiteit Leuven for osting is visit in Belgium wile on a sabbatical leave. Please do not quote witout autors permission. Corresponding autor s address: e.valdez@unsw.edu.au. 1

2 1 Introduction Consider an insurance company tat over some well-defined time reference, faces a random loss of X on a well-defined probability space (Ω, F, Pr). We can sometimes appropriately call X te risk faced by te insurer. For tis random loss X, assume tat its distribution function is denoted by F X (x) and its corresponding tail probability by F X (x) =1 F X (x). Assuming X is a continuous random variable, as is often te case, ten te density function of X is given by f X (x) = df X(x) dx = df X(x). dx Te task of te insurer is to assign a price tag on tis risk X leading us to examine premium principles. As pointed out by Young (2004), tat loosely speaking, a premium principle is a rule for assigning a premium to an insurance risk. Te insurance premium is usually coined risk-adjusted to refer to te fact tat it already incorporates model risks as well as parameter uncertainties. As advocated in tis paper, tese model and parameter risks are often andled by distorting te real probability measure. Tis paper introduces yet anoter metod to transform distributions. A premium principle π is a mapping from a set Γ of real-valued random variables defined on (Ω, F) to te set R of real numbers. Effectively, we ave π : Γ R so tat for every X belonging to Γ, π[x] R, is te assigned premium. To no surprise, premium principles are well-studied in te actuarial literature. See for example, Goovaerts, DeVijlder and Haezendonck (1984), Kaas, van Heerwaarden and Goovaerts (1994), and more recently Young (2004). A capter on insurance premium principles is devoted in Kaas, Goovaerts, Daene and Denuit (2001). Risk-adjusted premiums are often computed based on te expectation wit respect to a transformed probability measure, say Q, suctat π[x] =E Q [X] =E [ΨX] (1) were Ψ is a positive random variable wic tecnically is also te Radon-Nikodym derivative of te Q-measure wit respect to te real probability measure, sometimes denoted by P. See Gerber and Pafumi (1998). Tis is also sometimes called, in te Finance literature, te pricing density. In several premium principles, te Ψ follows te general form defined by a relation (X; λ) Ψ = (2) E [(X; λ)] for some function of te random variable X and a parameter λ often describing some level of risk-aversion to te insurer. Tis relation in (2) can also be justified in a decision teoretic framework as discussed in Heilman (1985) and Hürlimann (2004). Following Hürlimann (2004), for example, consider a loss function L : R 2 R wic describes te loss incurred L(x, x ) for a realized outcome X = x wen te action taken by te decision maker as been x. Decisions on premiums are determined on te basis of minimizing te expectation of tis loss function. Consider te loss function defined by L(x, x )=(x; λ) (x x ) 2. (3) 2

3 It is straigtforward to sow ten tat te resulting premium principle follows te general form (X; λ) π[x] =E E [(X; λ)] X. Many familiar premium principles lead to tis form. For example, te net premium principle can be derived wit (X; λ) =1and te variance premium principle can be derived wit (X; λ) =1+λX. Te celebrated Esscer premium principle can be derived by coosing (X; λ) =exp(λx), and te Karlsrue premium principle discussed in Hürlimann (2004) can be derived by coosing (X; λ) =X. Even te conditional tail expectation used as a risk measure can be derived wit (X; λ) =I X>FX 1 (q) were I( ) is an indicator function and FX 1 (q) is te quantile of te distribution of X at a pre-determined probability level q. In te univariate case, we ave X belonging to te class of elliptical distributions if its density can be expressed as " µx f X (x) = C # µ 2 σ g (4) σ for some so-called density generator g (wic is a function of non-negative variables), and were C is a normalizing constant. Later in te paper, te definition of elliptical distributions is expressed in terms of its caracteristic function because tis function always exist for a distribution. For te familiar Normal distribution N µ, σ 2 wit mean µ and variance σ 2, it is straigtforward to sow tat its density generator for wic we sall denote by g N in tis paper as te form g N (u) =exp( u/2). (5) In tis paper, we define elliptical transformations by considering te functional (x; λ) to ave te following form: Φ g 1 F X (x) + λ i 2 (x; λ) = Φ g 1 N F X (x) i 2 (6) were g is te density generator of some member of te class of elliptical distributions. Te family of elliptical distributions consists of symmetric distributions for wic te Student-t, te Uniform, Logistic, and Exponential Power distributions are oter familiar examples. Tis is a ric family of distributions tat allow for a greater flexibility in modelling risks tat exibit tails eavier tan tat of a Normal distribution. Tis can be especially useful for capturing insurance losses for extreme events. As we sall observe in our discussion in te paper, te probability transform defined using te ratio of density generators in (6) effectively distort te distribution by putting eavier weigts on bot tails of te distribution. In insurance premium calculations, wat tis translates to is giving more penalty on large losses but at te same time, placing some empasis on small or zero losses tereby encouraging small claims. Te result is terefore a trade-off between small and large claims redistributing te claims distribution so tat it leads to a more sensible and equitable 3

4 premium calculation. Tis trade-off is best parameterized by te parameter λ wic also effectively introduces a sift or translation of te distribution. Te rest of te paper as been organized as follows. In Section 2, we brieflydiscuss te ric class of elliptical distributions in te univariate dimension, togeter wit some important properties. Section 3 describes te notion of elliptical transformation. Several examples follow. In Section 4, we discuss te special case were te risk as a distribution tat is location-scale. Elliptical distributions are special cases of locationscale families. In Section 5, using some claims experience data, we demonstrate practical implementation of introducing premium loadings using te ideas developed in te paper. Finally, in section 6, we provide some concluding remarks togeter wit some remarks about possible direction for furter work. 2 Te Family of Elliptical Distributions Te class of elliptical loss distribution models provides a generalization of te class of normal loss models. In te following, we will describe tis class of models first in te univariate dimension, and briefly extending it to te multivariate dimension. Te class of elliptical distributions as been introduced in te statistical literature by Kelker (1970) and widely discussed in Fang, Kotz and Ng (1990). See also Landsman and Valdez (2003), Valdez and Daene (2003), and Valdez and Cerni (2003) for applications in insurance and actuarial science. Embrects, et al. (2001) also provides a fair amount of discussion of tis important class as a tool for modelling risk dependencies. 2.1 Definition of Elliptical Distributions It is widely known tat a random variable X wit a normal distribution as te caracteristic function expressed as E[exp(itX)] = exp(itµ) exp 1 2 t2 σ 2, (7) were µ and σ 2 are respectively, te mean and variance of te distribution. We sall use te notation X N µ, σ 2. Te class of elliptical distributions is a natural extension to te class of normal distributions. Definition 1 Te random variable X is said to ave an elliptical distribution wit parameters µ and σ 2 if its caracteristic function can be expressed as E [exp(itx)] = exp(itµ) ψ t 2 σ 2 (8) forsomescalarfunctionψ. If X as te elliptical distribution as defined above, we sall conveniently write X E µ, σ 2,ψ and say tat X is elliptical. Te function ψ is called te caracteristic generator of X and terefore, for te normal distribution, te caracteristic generator is clearly given by ψ(u) =exp( u/2). It is well-known tat te caracteristic function of a random variable always exists and tat tere is a one-to-one correspondence between distribution functions and caracteristic functions. Note owever tat not every function ψ can be used to construct a caracteristic function of an elliptical distribution. Obviously, tis function 4

5 ψ sould fulfill te requirement tat ψ(0) = 1. A necessary and sufficient condition for te function ψ to be a caracteristic generator of an ellipticial distribution can be seen in Teorem 2.2 of Fang, Kotz and Ng (1990). It is also interesting to note tat te class of elliptical distributions consists mainly of te class of symmetric distributions wic include well-known distributions like normal and Student-t. Te moments of X E µ, σ 2,ψ do not necessarily exist. However, it can be sown tat if te mean, E(X), exists, ten it will be given by E(X) =µ (9) andiftevariance,var(x), exists, ten it will be given by Var(X) = 2ψ 0 (0) σ 2, (10) were ψ 0 denotes te first derivative of te caracteristic generator. A necessary condition for te variance to exist is ψ 0 (0) <, (11) see Cambanis, Huang and Simmons (1981). In te case were µ =0and σ 2 =1, we ave wat we call a sperical distribution and te random variable X is replaced by a standard random variable Z. Tat is, we ave Z E(0, 1,ψ) and te notation S(ψ) for E(0, 1,ψ) is more typically used and tus, we write Z S(ψ). It is clear tat te caracteristic function of Z as te form E[exp(itZ)] = ψ t 2 for any real number t. Also, if we consider any random variable X satisfying X d = µ + σz, for some real number µ, some positive real number σ and some sperical random variable Z S(ψ), ten it can be sown tat X E µ, σ 2,ψ. Similarly, for any elliptical random variable X E µ, σ 2,ψ,wecanalwaysdefine te random variable Z = X µ σ wic is clearly sperical. 2.2 Densities of Elliptical Distributions An elliptically distributed random variable X E µ, σ 2,ψ does not necessarily possess a density function f X (x). In te case of a normal random variable X N µ, σ 2, its density is well-known to be " f X (x) = 1 exp 1 µ # x µ 2. (12) σ 2 σ 5

6 For elliptical distributions, one can prove tat if X E µ, σ 2,ψ as a density, it will ave te form " µx f X (x) = C # µ 2 σ g (13) σ for some non-negative function g( ) satisfying te condition 0 < Z 0 z 1/2 g(z) dz < (14) and a normalizing constant C given by Z C = z 1/2 g(z) dz 1. (15) 0 Also, te opposite statement olds: Any non-negative function g( ) satisfying te condition (14) can be used to define a one-dimensional density " µx # C µ 2 σ g σ of an elliptical distribution, wit C given by (15). Te function g( ) is called te density generator. One sometimes writes X E µ, σ 2,g for te one-dimensional elliptical distributions generated from te function g( ). A detailed proof of tese resultsfortecaseofn-dimension, using sperical transformations of rectangular coordinates, can be found in Landsman and Valdez (2003). From (12), one immediately finds tat te density generators and te corresponding normalizing constants of te normal random variable X N µ, σ 2 are given by g(z) =exp( z/2) and C = 1, respectively. Table 1 Some known elliptical distributions wit teir density generators Family Density generators g(u) Bessel g(u) =(u/b) a/2 K a (u/b) 1/2i, a> 1/2,b>0 were K a ( ) is te modified Bessel function of te 3rd kind Caucy g(u) =(1+u) 1 Exponential Power g(u) =exp[ r (u) s ], r, s > 0 Laplace g(u) =exp( u ) Logistic g(u) = Normal Student-t g(u) = exp( u) [1 + exp( u)] 2 g N (u) =exp( u/2) ³ 1+ u m (m+1)/2, m>0 an integer 6

7 As a special case, we find tat from te results concerning elliptical distributions, if a sperical random variable Z S(ψ) possess a density f Z (z), tenitwillave te form f Z (z) =Cg z 2, were te density generator g satisfies te condition (14) and te normalizing constant C satisfies (15). Furtermore, te opposite also olds: any non-negative function g( ) satisfying te condition (14) can be used to define a one-dimensional density Cg z 2 of a sperical distribution wit te normalizing constant C satisfying (15). One often writes S(g) for te sperical distribution generated from te density generator g( ). 3 Elliptical Transforms Suppose g Z (u) is te density generator of a univariate sperical random variable Z E (0, 1,g Z ), also written as Z S (g Z ). Weavesubscriptedtedensity generator to empasize te corresponding sperical random variable Z. Nowwe define te ratio of density generators as follows. Definition 2 Let X be a random variable wit tail probability function F X ( ). We define te ratio of density generators g Z and g N to be te random variable Φ g 1 Z F X (X) + λ i 2 gz (X; λ) = Φ g 1 N F X (X) i 2 (16) for some non-negative parameter λ 0 and were Φ ( ) is te cumulative distribution function of a standard Normal random variable and g N is te density generator of a Normal distribution. Notice tat in te definition we assume we ave a random variable X Γ, belonging to te set of all risks, wose tail probability function is given by F X ( ). As we sall observe later in te examples tat follow, tis non-negative parameter can be interpreted as a premium per unit of volatility (or risk). Te following teorem gives an expression for te expectation of te ratio of density generators. Teorem 1 Let X be a random variable wit tail probability function F X ( ) and let te cumulative distribution function of a standard Normal be denoted by Φ ( ). Te expectation of te ratio of density generators g Z and g N can be expressed as E [ gz (X; λ)] = 1 Z g Z z 2 dz. (17) Proof. First assuming te density of X is denoted by f X ( ), ten we can write te expectation as Z Φ g 1 Z F X (x) + λ i 2 E[ gz (X; λ)] = Φ g 1 N F X (x) i 2 f X(x) dx. 7

8 Using te transformation u = F X (x) so tat du = f X (x) dx, tenweave Z Φ 1 g 1 Z (u)+λ i 2 E[ gz (X; λ)] = du. 0 g N (Φ 1 (u)) 2i Wit anoter transformation w = Φ 1 (u) so tat du = 1 g N w 2 dw, wetenave Z g Z (w + λ) 2i 1 E[ gz (X; λ)] = g N (w 2 g N w 2 dw. ) Applying one last transformation z = w + λ, we get te desired result. Following te above Teorem, it is clear tat for a sperical Z S (g Z ) wit a normalizing constant C Z, we ave te following result (wic will later be useful): 1 Z C Z = g Z z 2 dz 1 = 1 E[gZ (X; λ)]. We now give some examples of ratios of density generators. Example 3.1: Normal-to-Normal Generators If g Z is also te density generator of a Normal distribution, ten we ave Φ g 1 N F X (X) + λ i 2 gn (X; λ) = Φ g 1 N F X (X) i 2 =exp λ Φ 1 F X (X) λ = e λ2 /2 exp λφ 1 F X (X). (18) It is easy to see tat in tis case, te expectation of (18) is equal to 1. application of Teorem 1, for example, leads us to tis. Direct Example 3.2: Student-t-to-Normal Generators If g Z is te density generator of a Student-t distribution wit m degrees of freedom, ten we ave 2 Φ g 1 Z F X (X) + λ i 2 gz (X; λ) = Φ g 1 N F X (X) i 2 = 1+ 1 m Φ 1 F X (X) + λ i 2 (m+1)/2 exp 12 Φ 1 F X (X) i 2. (19) Applying Teorem 1 to solve for te expectation, we ave E[ gz (X; λ)] = = Z 1 Z 1 1 correction made: 1-May correction made: 1-May g Z z 2 dz µ 1+ 1 m z2 (m+1)/2 dz. 8

9 Applying transformations, it can be sown tat tis expectation leads us to r Z m 1 r µ E[ gz (X; λ)] = z (m/2) 1 (1 z) (1/2) 1 m m dz = 0 B 2, 1 2 r r m Γ(m/2) Γ(1/2) m = Γ((m +1)/2) = Γ(m/2) 2 Γ((m +1)/2), were B(, ) is te Beta function and Γ( ) is te Gamma function. Here, we also note tat we applied te fact tat Γ(1/2) = π. Now interestingly, te case were m =1 leads us to te Caucy-to-Normal generators and in wic case, we would ave r r 1 Γ(1/2) Γ(1/2) π E[ gz (X; λ)] = = Γ(1) 2. It can also be sown tat in te limiting case were m, tis leads us back to te Normal-to-Normal density generators. Example 3.3: Exponential Power-to-Normal Generators Assuming now tat g Z is te density generator of an Exponential-Power distribution (see Table 1), ten we ave Φ g 1 Z F X (X) + λ i 2 gz (X; λ) = Φ g 1 N F X (X) i 2 = n exp r Φ 1 F X (X) + λ 2s 1 2 Φ 1 F X (X) io 2. (20) Applying Teorem 1 to solve for te expectation, we ave E[ gz (X; λ)] = = = Z 1 Z 1 g Z z 2 dz e rz2s dz Z 2 e rz2s dz 0 were te tird line follows because of symmetry. Let us derive an explicit form for tis expectation. First, consider te transformation u = z 2s so tat du =2sz 2s 1 dz = 2su (2s 1)/2s dz. Tis leads us to E[ gz (X; λ)] = 1 Z s u ( 2s) 1 1 e ru du. 0 Next ten, consider te transformation z = ru so tat dz = rdu. Weave Z 1 E[ gz (X; λ)] = s r1 (1/2s) z ( 2s) 1 1 e z dz 0 µ 1 1 = s r1 (1/2s) Γ, 2s were Γ( ) is te usual Gamma function. Te case were r =1/2 and s =1leads us to te Normal distribution case. 9

10 Consider again g Z as te density generator of a univariate sperical random variable Z S(g Z ).Nowwedefine wat we meant by probability transformation using tese elliptical density generators. Tese transformations we will, for simplicity, call elliptical transformations. Definition 3 Let X be a random variable wit tail probability function F X ( ) and wose density function exists and is equal to f X ( ). Wedefine te transformed random variable, denoted by X, to be one wit a (transformed) density function given by Φ g 1 Z F X (X) + λ i 2 f X (x) =C Φ g 1 N F X (X) i 2 f X(x) (21) for all x in te domain or range of X and were C is a normalizing constant. By recognizing tat gz (X; λ) = g Z (Φ 1 (F X (X))+λ) 2i g N (Φ 1 (F X (X))) 2i, te ratio of density generators, we simply note tat te normalizing constant is C = 1 E[ gz (X; λ)] for wic can be easily evaluated from Teorem 1. We can, as a matter of fact, find an expression for te distribution function of te (transformed) random variable X. First, notice tat Z Z Φ g 1 Z F X (v) + λ i 2 F X (x) = f X (v) dv = C Φ x x g 1 N F X (v) i 2 f X(v) dv. Applying te transformation u = F X (v) so tat du = f X (v) dv, tis yields Z Φ F X (x) g 1 Z (u)+λ i 2 F X (x) =C du 0 g N (Φ 1 (u)) 2i and ten applying te transformation w = Φ 1 (u) so tat du = 1 g N w 2 dw, we ave F X (x) = = Z C Φ 1 (F X (x)) Z Φ 1 (F X (x))+λ g Z (w + λ) 2i g N (w 2 ) C Z g Z z 2 dz g N w 2 dw = F Z Φ 1 F X (x) + λ (22) were F Z ( ) is te distribution function of a sperical random variable wit density generator g Z. In sort, we ave Z S(g Z ). 10

11 In an independent work done by Landsman (2004), e also proposed to use a transformation of densities of elliptical random variables using a ratio of te density generator evaluated as i g ((x µ) /σ) 2 2λx g ((x µ) /σ) 2i. Landsman (2004) sowed tat tis transformation is a generalization of te Esscer transform for elliptical distributions. He furter demonstrated tat tis generalizes te variance premium principle applied again to elliptical distributions and e appropriately called tese transformations elliptical tilting. Tere are several differences between Landsman s elliptical tilting wit tose proposed in tis paper. First, unlike Landsman (2004), we take te ratio of two different density generators, one of wic is always te Normal density generator, and te oter is an arbitrarily cosen one. In effect, we are transforming te distribution by taking te relative weigts of te densities of an arbitrarily cosen elliptical random variable to te Normal random variable. Te arbitrary selection allows te decision maker (wic in tis case is te insurer) some flexibility. Landsman cooses te density generator of te random variable being tilted wic is always an elliptical random variable. Second, wile bot Landsman and tis paper allows translation of te distribution by introducing a sift parameter λ, te manner in wic te translation is accomplised differ in bot respects. Landsman introduces te sift using ((x µ) /σ) 2 2λx wile we, in tis paper, introduce te sift via Φ 1 F X (x) + λ. Tird, Landsman limited its applications to exponential tilting of elliptical random variables wile ours does not necessarily ave suc limitations. Lastly, in te case of a Normal distribution, Landsman generalizes te variance premium principle wile we generalize te standard deviation premium principle. Te standard deviation is more commonly utilized as a measure of te level of riskiness of a portfolio. See, for example, Markowitz (1952) and Merton (1990). Definition 4 Let X be a random variable wit tail probability function F X ( ) and wose density function exists and is equal to f X ( ). LetX be te transformed random variable of X according to te elliptical transformation defined in (21). Ten te expectation of X is defined to be te premium principle implied by te elliptical transformation: π[x] =E(X gz (X; λ) )=E E[ gz (X; λ)] X. Observe tat from (22), we can also derive te premium (or expectation of te transformed distribution) using π[x] =E(X )= Z 0 F Z Φ 1 F X (x) + λ dx + Z 0 F Z Φ 1 F X (x) + λ dx wic reduces to just R 0 F Z Φ 1 F X (x) + λ dx for random variables wit nonnegative support. We produce Figures 1 to 4 to elp us visualize te transformation. All figures in tis paper are attaced as appendix following te list of references. For Figures 11

12 1 troug 4, we assume te random variable X being transformed is Uniform on te interval (0, 100). Tis coice as been made to elp us more vividly visualize te effects of te transformation. Te straigt orizontal broken line found in tese figures is te un-transformed (or original) Uniform density. In Figure 1, we consider te case were we coose Z to be te Normal distribution and we present te various transformation for different values of λ. Generally, we see tat using te Normal density generator, tis puts more weigts on te rigttail of te distribution. So if te risk are losses as in te case of insurances, tis penalizes more eavily large amount of losses. Te effect of λ introduces a sift of te distribution to te rigt for larger λ puttinginaneveneavierpenaltyonte tails. In Figures 2, 3 and 4, we consider te case were we coose Z to be te Student-t distribution. Figure 2 examines te effect of increasing te degrees of freedom (fixing λ =1) wile Figure 3 examines te effect of increasing te risk aversion parameter λ (fixing m =5). In Figure 4, we present te case were λ =0wen tere is no sift in te distribution. Here we observe equal penalty of bot ends of te tail. Te Student-t distortion as been first introduced by Wang (2004) in te pricing of catastrope bonds. As pointed out in is paper, putting more weigts to te tails of te underlying distributions is a way to reflect investor beavior and preferences tat we often observe in te market. Investors generally fear large unexpected losses, but tey also like large unexpected gains. Interestingly as special cases, we recover some of te familiar premium principles as discussed below. Tis in some sense generalizes tese premium principles. Example 3.4: Wang Premium Principle If g Z is cosen to be te density generator of a Normal distribution, ten it is immediate to get te Wang transformation: F X (x) =Φ Φ 1 F X (x) + λ. See Wang (1996) and Wang (2000). Example 3.5: Wang s Student-t Distortion Premium Principle If g Z is cosen to be te density generator of a Student-t distribution wit m degrees of freedom, ten we ave as defined in Wang (2004) F X (x) =Q Φ 1 F X (x) + λ, were, following Wang s notation, Q( ) denotes te distribution function of a Studenttwitm degrees of freedom. Wang actually as set λ =0and used F X (x) = Q Φ 1 F X (x) instead. Example 3.6: Esscer Premium Principle Because 1 Φ(x) =Φ( x), ten in te special case were X is Normally distributed say N µ, σ 2, it is rater straigtforward to see tat te elliptical transformation leads to te Esscer transform. In tis case, we ave gz (X; λ) E[ gz (X; λ)] = e λ2/2 exp λφ 1 F X (X) E e λ2 /2 exp λφ 1 F X (X) i = exp λσ X E exp λ σ X. See Esscer (1932) and also more recent articles, for example, Gerber and Siu (1994). 12

13 4 Location-Scale Families By considering te risk X to belong to families of location-scale distributions, we can furter simplify te premium principle implied by te elliptical transformation. Members of te elliptical distributions also belong to te larger family of locationscale distributions. As a matter of fact, in tis case, we are able to recover a similar concept to te standard deviation premium principle. We give te following teorem. Teorem 2 Let µ X be a random variable belonging to a location-scale family so tat x µ F X (x) =F Z for some Z = X µ wose distribution is independent of σ σ te location parameter µ and scale parameter σ. Ten te premium principle implied by te elliptical transformation can be expressed as i π[x] =µ + E Z F 1 Z (Φ(Z λ)) σ. Proof. First assuming te density of X exists and is denoted by f X ( ), tenwe can write te expectation as Z π[x] =C x Φ g 1 Z F X (x) + λ i 2 Φ g 1 N F X (x) i 2 f X (x) dx. Applying te transformation u = F X (x) so tat du = f X (x) dx, tis yields to Z Φ 1 π[x] =C F 1 g 1 Z (u)+λ i 2 X (u) du. 0 g N (Φ 1 (u)) 2i Wit anoter transformation w = Φ 1 (u) so tat du = 1 g N w 2 dw, wetenave Z π[x] =C F 1 g Z (w + λ) 2i 1 X (Φ(w)) g N (w 2 g N w 2 dw. ) Applying one last transformation z = w + λ, weget π[x] = Z F 1 X (Φ(z λ)) C Z g Z z 2 dz R were C Z = g Z z 2 i 1 dz is te normalizing constant of te sperical random variable Z. Tus, we see tat i π[x] =E Z F 1 X (Φ(Z λ)) were te expectation E Z is evaluated wit respect to te sperical random variable Z. Now, since X is location scale, we can re-write F 1 X (Φ(Z λ)) = µ + F 1 Z (Φ(Z λ)) σ and te desired result immediately follows. 13

14 Tis result gives us a way of interpreting te parameter λ, wic is actually a risk premium per unit of risk (or volatility) as measured by te standard deviation. IntecasewereX is normally distributed N µ, σ 2 and we coose Z to be te standard Normal, ten we get i E Z F 1 Z (Φ(Z λ)) = E Z (λ Z) =λ E Z (Z) =λ, so tat te resulting premium principle leads us to te familiar form of te standard deviation premium principle: π[x] =µ + λσ. Terefore, we ave λ = π[x] µ, a risk premium per unit of risk as measured σ by σ. Furtermore, it can also be sown tat by using a first-order Taylor s series expansion, we can ave te following approximation: E Z F 1 Z (Φ(Z λ)) i F 1 1 Z (Φ( λ)) = F Z Φ(λ), were Φ denotes te tail probability function of a standard normal. 5 Practical Implementation In tis section, we illustrate ow to practically implement te premium principles developed in tis paper using te empirical losses investigated in Frees and Valdez (1998). Te experience consisted of a random sample of 1,500 claims from a general insurance liability portfolio provided by te Insurance Services Office, Inc. For our purposes, we only include in tis analysis te observations from claims, for wic we denote ere as te risk X. Unfortunately, te experience data we consider ere consists of observations tat generated claims arising from an insurance portfolio. In an ordinary insurance portfolio, te case is usually one were tere is a mass of zero claims, some observations tat never generated claims. In te pricing of insurance, one would take into account te probability mass of zero claims. For te lack of available insurance claims data, we resort to tis experience data, owever, we ask te reader to exercise caution tat normally in insurance pricing, te probability mass of zero claims must be taken into account. Tis mass as been ignored in te ensuing analysis. Summary statistics can be found in Frees and Valdez (1998), but just to reiterate tese statistics, te average loss was 41,208, te median was 12,000 and te standard deviation was 102,748. In fitting for te best distribution for tis experience loss data, censoring ad to be considered because claims ave policy limits and if te actual loss exceeded te policy limit, only te policy limit was recorded. At any rate, according to Frees and Valdez (1998), te best fitting loss distribution model for te data as te Pareto form described by µ λ θ F X (x) =, for x>0. λ + x Its density function terefore as te form µ 1 θ+1 f X (x) =θλ θ. λ + x 14

15 Te maximum likeliood parameter estimates are b λ =14, 453 and b θ =1.135, wit respective standard errors of 1,397 and In Figure 5, we provide a frequency istogram of te logaritm of te observed losses (called logloss) wit te fitted Pareto density function of te logloss. Tis figure ignores te censored observations. For simplicity, in constructing te premium principle used to illustrate, we ave considered te coice of a Student-t density generator wit m =3degrees of freedom. Tere is no empirical basis for tis coice, except te Student-t appears to be te most sensible as it allows flexibility wit te additional parameter for degrees of freedom. Tus, from Example 3.2, we can write 3 gz (X; λ) = " exp " µ µ ³ # Φ 1 λ θ 2 2 λ+x + λ 1 2 µ µ ³ θ # 2. Φ 1 λ λ+x Te purpose ere is terefore to develop risk-adjusted premiums. Te results are summarized in Figures 6 and 7. Figure 6 provides te risk-adjusted premiums for varying values of λ and it is not surprising to see tat te risk-adjusted premiums increase wit increasing λ. For te different risk-adjusted premiums implied by te different λ, we computed ten te probability of aving losses smaller tan tese riskadjusted premiums, under te original probability measure. Tis level of probability we denote ere by q, and in Figure 7, we sow te risk-adjusted premiums as a function of tis probability level. Again, tere is iger risk-adjusted premiums associated wit larger q. Notice tat te risk-adjusted premium does not always yield a larger value tan te unadjusted premium (tat is, te net premium equal to λ/ (θ 1) in te Pareto case), but because of its increasing nature, for some level λ and correspondingly for some level q, beyond wic tere will be a positive risk premium. Numerically, tese results are summarized as Table 2 for convenience. Here, te net premium level is 107,059, and at λ = (or correspondingly q =0.9108), tis premium level will equal te risk-adjusted premium. See Table 2. Table 2 Te risk-adjusted premiums and teir probability levels λ unadjusted premium E(X) risk-adjusted premium π[x] probability level q ,059 4, ,059 11, ,059 31, ,059 91, , , ,059 1,051, ,059 4,462, correction made: 1-May

16 6 Final Remarks Tis paper introduces te notion of elliptical transformation leading to a premium principle wic in some sense generalizes te familiar Wang transformation as well as te premium principle introduced by Wang (2004) using te Student-t distortion function. We also note tat in te special case of transforming random variables belonging to te location-scale families, te resulting premium principle is te standard deviation premium principle. Furtermore, we observe from te figures presented in te paper tat in some sense, te transformation introduces a eavy penalty on te extreme rigt tails of te distribution but also encourages small losses by placing relatively reasonable weigts on te extreme left tails of te distribution. How muc tis penalty is can depend upon te coice of te density generator togeter wit te parameter λ wic in a sense gives a measure of aversion to te level of risk of te insurer. Tis parameter also introduces a sift in te distribution of te risk. In te Normal distribution case, te same parameter leads to an interpretation of a risk premium per unit of volatility, or risk. We are also able to sow tat te elliptical transformation recovers many oter familiar premium principles. Te notion of elliptical transformation introduced in tis paper can also be applied as a risk measure to compute economic capital. It will be interesting ow tis risk measure may be extended to te case were tere is a need to aggregate several risks or to re-allocate te total economic capital into various constituents. Moreover, in te future, it will be an interesting work to examine some properties of tis premium principle. References [1] Cambanis, S., Huang, S., and Simons, G. (1981) On te Teory of Elliptically Contoured Distributions, Journal of Multivariate Analysis 11: [2] Embrects, P., McNeil, A., Straumann, D. (2001). Correlation and Dependence in Risk Management: Properties and Pitfalls. Risk Management: Value at Risk and Beyond, edited by Dempster, M. and Moffatt, H.K. Cambridge University Press. Also available at < [3] Esscer, F. (1932) On te Probability Function in te Collective Teory of Risk, Skandinavisk Aktuarietidskrift 15: [4] Fang, K.T., Kotz, S. and Ng, K.W. (1987) Symmetric Multivariate and Related Distributions. London: Capman & Hall. [5] Frees, E.W. and Valdez, E.A. (1998) Understanding Relationsips Using Copulas, Nort American Actuarial Journal 2: [6] Gerber, H.U. and Pafumi, G. (1998) Utility Functions: from Risk Teory to Finance, Nort American Actuarial Journal 2: [7] Gerber, H.U. and Siu, E. (1994) Option Pricing by Esscer Transforms, Transactions of te Society of Actuaries XLVI: [8] Goovaerts, M., DeVijlder, F. and Haezendonck, J. (1984) Insurance Premiums. Amsterdam: Nort Holland Publisers. 16

17 [9] Heilman, W.R. (1987) A Premium Calculation Principle for Large Risks, in Operations Researc Proceedings 1986, pp Berlin-Heidelberg: Springer Verlag. [10] Hürlimann, W.(2004) Is te Karlsrue Premium a Fair Value Premium? Blätter der Deutscen Gesellscaft für Versiceringsmatematik XXXVI(4): [11] Kaas, R., Goovaerts, M., Daene, J. and Denuit, M. (2001) Modern Actuarial Risk Teory. Neterlands: Kluwer Academic Publisers. [12] Kaas, R., van Heerwaarden, A.E. and Goovaerts, M. (1994) Ordering of Actuarial Risks. Amsterdam: Institute for Actuarial Science and Econometrics, University of Amsterdam. [13] Kelker, D. (1970). Distribution Teory of Sperical Distributions and Location- Scale Parameter Generalization. Sankya 32: [14] Landsman, Z. (2004) On te Generalization of Esscer and Variance Premiums Modified for te Elliptical Family of Distributions, Insurance: Matematics & Economics 35: [15] Landsman, Z. and Valdez, E.A. (2003) Tail Conditional Expectations for Elliptical Distributions, Nort American Actuarial Journal 7: [16] Markowitz, H. (1952) Portfolio Selection, Journal of Finance 7: [17] Merton, R.C. (1990) Continuous Time Finance. Massacusetts: Blackwell Publising Ltd. [18] Panjer, H.H. (2002) Measurement of Risk, Solvency Requirements, and Allocation of Capital witin Financial Conglomerates, Institute of Insurance and Pension Researc, University of Waterloo, Researc Report [19] Valdez, E.A. and Cerni, A. (2003) Wang s Capital Allocation Formula for Elliptically Contoured Distributions, Insurance: Matematics & Economics 33: [20] Valdez, E.A. and Daene, J. (2003) Bounds for Sums of Non-Independent Log-Elliptical Random Variables, paper presented at te Sevent International Congress on Insurance: Matematics & Economics, Lyon, France. [21] Wang, S.S. (1996) Premium Calculation by Transforming te Layer Premium Density, ASTIN Bulletin 21: [22] Wang, S.S. (2000) A Class of Distortion Operators for Pricing Financial and Insurance Risks, Journal of Risk and Insurance 67: [23] Wang, S.S. (2004) Cat Bond Pricing Using Probability Transforms, invited article, Geneva Pa-pers: Etudes et Dossiers, special issue on Insurance and te State of te Art in Cat Bond Pricing, 278: [24] Young, V.R. (2004) Premium Principles in Encyclopedia of Actuarial Science, ed. Sundt, B. and Teugels, J., New York: Jon Wiley & Sons, Ltd. 17

18 Transformed Densities lam bda=1 lam bda=2 lam bda=3 original x Figure 1: Elliptical transformation using te Normal density generator. Tis transformation effectively results in te Wang transformation sowing large penalty on te rigt tail. Transformed Densities m=1 m=5 m=10 original x Figure 2: Elliptical transformation using te Student-t density generator, varying te degrees of freedom. Tis transformation effectively penalizes bot extremes of te distribution, but more so wit rigt tail. 18

19 Transformed Densities lambda=1 lambda=3 lambda=5 original x Figure 3: Elliptical transformation still using te Student-t density generator, but varying te parameter λ. Here we ave fixed te degrees of freedom to 5. Transformed Densities x Figure 4: Elliptical transformation using te Student-t density generator. Tisistecasewereλ =0and degrees of freedom m =5. It sows tat equal penalties are imposed at extremes of te distribution. 19

20 Pareto Fit LOGLOSS Figure 5: Frequency istogram of te Logaritm of claims. Te smoot curve superimposed on te istogram is te fitted Pareto distribution. Risk-Adjusted Premiums lambda Figure 6: Te risk-adjusted premium as a function of te distribution sift parameter λ. Te broken orizontal line gives te unadjusted net premium, tat is, te expected value of te un-transformed distribution. A positive loading terefore results only wen λ is cosen so tat te smoot curve is above te orizontal line. 20

21 Risk-Adjusted Premiums probability (q) Figure 7: Te risk-adjusted premium as a function of te level of probability q. Te broken orizontal line gives te unadjusted net premium, tat is, te expected value of te un-transformed distribution. 21

Probability Transforms with Elliptical Generators

Probability Transforms with Elliptical Generators Probability Transforms with Elliptical Generators Emiliano A. Valdez, PhD, FSA, FIAA School of Actuarial Studies Faculty of Commerce and Economics University of New South Wales Sydney, AUSTRALIA Zurich,

More information

Tail Conditional Expectations for Extended Exponential Dispersion Models

Tail Conditional Expectations for Extended Exponential Dispersion Models American Researc Journal of Matematics Original Article ISSN 378-704 Volume 1 Issue 4 015 Tail Conditional Expectations for Extended Exponential Dispersion Models Ye (Zoe) Ye Qiang Wu and Don Hong 1 Program

More information

Volume 29, Issue 3. Existence of competitive equilibrium in economies with multi-member households

Volume 29, Issue 3. Existence of competitive equilibrium in economies with multi-member households Volume 29, Issue 3 Existence of competitive equilibrium in economies wit multi-member ouseolds Noriisa Sato Graduate Scool of Economics, Waseda University Abstract Tis paper focuses on te existence of

More information

Some results on Denault s capital allocation rule

Some results on Denault s capital allocation rule Faculty of Economics and Applied Economics Some results on Denault s capital allocation rule Steven Vanduffel and Jan Dhaene DEPARTMENT OF ACCOUNTANCY, FINANCE AND INSURANCE (AFI) AFI 0601 Some Results

More information

Polynomial Interpolation

Polynomial Interpolation Capter 4 Polynomial Interpolation In tis capter, we consider te important problem of approximatinga function fx, wose values at a set of distinct points x, x, x,, x n are known, by a polynomial P x suc

More information

A = h w (1) Error Analysis Physics 141

A = h w (1) Error Analysis Physics 141 Introduction In all brances of pysical science and engineering one deals constantly wit numbers wic results more or less directly from experimental observations. Experimental observations always ave inaccuracies.

More information

New families of estimators and test statistics in log-linear models

New families of estimators and test statistics in log-linear models Journal of Multivariate Analysis 99 008 1590 1609 www.elsevier.com/locate/jmva ew families of estimators and test statistics in log-linear models irian Martín a,, Leandro Pardo b a Department of Statistics

More information

232 Calculus and Structures

232 Calculus and Structures 3 Calculus and Structures CHAPTER 17 JUSTIFICATION OF THE AREA AND SLOPE METHODS FOR EVALUATING BEAMS Calculus and Structures 33 Copyrigt Capter 17 JUSTIFICATION OF THE AREA AND SLOPE METHODS 17.1 THE

More information

Financial Econometrics Prof. Massimo Guidolin

Financial Econometrics Prof. Massimo Guidolin CLEFIN A.A. 2010/2011 Financial Econometrics Prof. Massimo Guidolin A Quick Review of Basic Estimation Metods 1. Were te OLS World Ends... Consider two time series 1: = { 1 2 } and 1: = { 1 2 }. At tis

More information

Polynomial Interpolation

Polynomial Interpolation Capter 4 Polynomial Interpolation In tis capter, we consider te important problem of approximating a function f(x, wose values at a set of distinct points x, x, x 2,,x n are known, by a polynomial P (x

More information

1 The concept of limits (p.217 p.229, p.242 p.249, p.255 p.256) 1.1 Limits Consider the function determined by the formula 3. x since at this point

1 The concept of limits (p.217 p.229, p.242 p.249, p.255 p.256) 1.1 Limits Consider the function determined by the formula 3. x since at this point MA00 Capter 6 Calculus and Basic Linear Algebra I Limits, Continuity and Differentiability Te concept of its (p.7 p.9, p.4 p.49, p.55 p.56). Limits Consider te function determined by te formula f Note

More information

Numerical Differentiation

Numerical Differentiation Numerical Differentiation Finite Difference Formulas for te first derivative (Using Taylor Expansion tecnique) (section 8.3.) Suppose tat f() = g() is a function of te variable, and tat as 0 te function

More information

. If lim. x 2 x 1. f(x+h) f(x)

. If lim. x 2 x 1. f(x+h) f(x) Review of Differential Calculus Wen te value of one variable y is uniquely determined by te value of anoter variable x, ten te relationsip between x and y is described by a function f tat assigns a value

More information

Taylor Series and the Mean Value Theorem of Derivatives

Taylor Series and the Mean Value Theorem of Derivatives 1 - Taylor Series and te Mean Value Teorem o Derivatives Te numerical solution o engineering and scientiic problems described by matematical models oten requires solving dierential equations. Dierential

More information

Exam 1 Review Solutions

Exam 1 Review Solutions Exam Review Solutions Please also review te old quizzes, and be sure tat you understand te omework problems. General notes: () Always give an algebraic reason for your answer (graps are not sufficient),

More information

Differentiation in higher dimensions

Differentiation in higher dimensions Capter 2 Differentiation in iger dimensions 2.1 Te Total Derivative Recall tat if f : R R is a 1-variable function, and a R, we say tat f is differentiable at x = a if and only if te ratio f(a+) f(a) tends

More information

1 Calculus. 1.1 Gradients and the Derivative. Q f(x+h) f(x)

1 Calculus. 1.1 Gradients and the Derivative. Q f(x+h) f(x) Calculus. Gradients and te Derivative Q f(x+) δy P T δx R f(x) 0 x x+ Let P (x, f(x)) and Q(x+, f(x+)) denote two points on te curve of te function y = f(x) and let R denote te point of intersection of

More information

The Priestley-Chao Estimator

The Priestley-Chao Estimator Te Priestley-Cao Estimator In tis section we will consider te Pristley-Cao estimator of te unknown regression function. It is assumed tat we ave a sample of observations (Y i, x i ), i = 1,..., n wic are

More information

THE IDEA OF DIFFERENTIABILITY FOR FUNCTIONS OF SEVERAL VARIABLES Math 225

THE IDEA OF DIFFERENTIABILITY FOR FUNCTIONS OF SEVERAL VARIABLES Math 225 THE IDEA OF DIFFERENTIABILITY FOR FUNCTIONS OF SEVERAL VARIABLES Mat 225 As we ave seen, te definition of derivative for a Mat 111 function g : R R and for acurveγ : R E n are te same, except for interpretation:

More information

The Laplace equation, cylindrically or spherically symmetric case

The Laplace equation, cylindrically or spherically symmetric case Numerisce Metoden II, 7 4, und Übungen, 7 5 Course Notes, Summer Term 7 Some material and exercises Te Laplace equation, cylindrically or sperically symmetric case Electric and gravitational potential,

More information

Lecture 15. Interpolation II. 2 Piecewise polynomial interpolation Hermite splines

Lecture 15. Interpolation II. 2 Piecewise polynomial interpolation Hermite splines Lecture 5 Interpolation II Introduction In te previous lecture we focused primarily on polynomial interpolation of a set of n points. A difficulty we observed is tat wen n is large, our polynomial as to

More information

Complexity of Decoding Positive-Rate Reed-Solomon Codes

Complexity of Decoding Positive-Rate Reed-Solomon Codes Complexity of Decoding Positive-Rate Reed-Solomon Codes Qi Ceng 1 and Daqing Wan 1 Scool of Computer Science Te University of Oklaoma Norman, OK73019 Email: qceng@cs.ou.edu Department of Matematics University

More information

Quantum Theory of the Atomic Nucleus

Quantum Theory of the Atomic Nucleus G. Gamow, ZP, 51, 204 1928 Quantum Teory of te tomic Nucleus G. Gamow (Received 1928) It as often been suggested tat non Coulomb attractive forces play a very important role inside atomic nuclei. We can

More information

2.8 The Derivative as a Function

2.8 The Derivative as a Function .8 Te Derivative as a Function Typically, we can find te derivative of a function f at many points of its domain: Definition. Suppose tat f is a function wic is differentiable at every point of an open

More information

f a h f a h h lim lim

f a h f a h h lim lim Te Derivative Te derivative of a function f at a (denoted f a) is f a if tis it exists. An alternative way of defining f a is f a x a fa fa fx fa x a Note tat te tangent line to te grap of f at te point

More information

(a) At what number x = a does f have a removable discontinuity? What value f(a) should be assigned to f at x = a in order to make f continuous at a?

(a) At what number x = a does f have a removable discontinuity? What value f(a) should be assigned to f at x = a in order to make f continuous at a? Solutions to Test 1 Fall 016 1pt 1. Te grap of a function f(x) is sown at rigt below. Part I. State te value of eac limit. If a limit is infinite, state weter it is or. If a limit does not exist (but is

More information

Consider a function f we ll specify which assumptions we need to make about it in a minute. Let us reformulate the integral. 1 f(x) dx.

Consider a function f we ll specify which assumptions we need to make about it in a minute. Let us reformulate the integral. 1 f(x) dx. Capter 2 Integrals as sums and derivatives as differences We now switc to te simplest metods for integrating or differentiating a function from its function samples. A careful study of Taylor expansions

More information

Symmetry Labeling of Molecular Energies

Symmetry Labeling of Molecular Energies Capter 7. Symmetry Labeling of Molecular Energies Notes: Most of te material presented in tis capter is taken from Bunker and Jensen 1998, Cap. 6, and Bunker and Jensen 2005, Cap. 7. 7.1 Hamiltonian Symmetry

More information

SFU UBC UNBC Uvic Calculus Challenge Examination June 5, 2008, 12:00 15:00

SFU UBC UNBC Uvic Calculus Challenge Examination June 5, 2008, 12:00 15:00 SFU UBC UNBC Uvic Calculus Callenge Eamination June 5, 008, :00 5:00 Host: SIMON FRASER UNIVERSITY First Name: Last Name: Scool: Student signature INSTRUCTIONS Sow all your work Full marks are given only

More information

Notes on wavefunctions II: momentum wavefunctions

Notes on wavefunctions II: momentum wavefunctions Notes on wavefunctions II: momentum wavefunctions and uncertainty Te state of a particle at any time is described by a wavefunction ψ(x). Tese wavefunction must cange wit time, since we know tat particles

More information

ch (for some fixed positive number c) reaching c

ch (for some fixed positive number c) reaching c GSTF Journal of Matematics Statistics and Operations Researc (JMSOR) Vol. No. September 05 DOI 0.60/s4086-05-000-z Nonlinear Piecewise-defined Difference Equations wit Reciprocal and Cubic Terms Ramadan

More information

SECTION 3.2: DERIVATIVE FUNCTIONS and DIFFERENTIABILITY

SECTION 3.2: DERIVATIVE FUNCTIONS and DIFFERENTIABILITY (Section 3.2: Derivative Functions and Differentiability) 3.2.1 SECTION 3.2: DERIVATIVE FUNCTIONS and DIFFERENTIABILITY LEARNING OBJECTIVES Know, understand, and apply te Limit Definition of te Derivative

More information

Regularized Regression

Regularized Regression Regularized Regression David M. Blei Columbia University December 5, 205 Modern regression problems are ig dimensional, wic means tat te number of covariates p is large. In practice statisticians regularize

More information

Math 102 TEST CHAPTERS 3 & 4 Solutions & Comments Fall 2006

Math 102 TEST CHAPTERS 3 & 4 Solutions & Comments Fall 2006 Mat 102 TEST CHAPTERS 3 & 4 Solutions & Comments Fall 2006 f(x+) f(x) 10 1. For f(x) = x 2 + 2x 5, find ))))))))) and simplify completely. NOTE: **f(x+) is NOT f(x)+! f(x+) f(x) (x+) 2 + 2(x+) 5 ( x 2

More information

Lecture XVII. Abstract We introduce the concept of directional derivative of a scalar function and discuss its relation with the gradient operator.

Lecture XVII. Abstract We introduce the concept of directional derivative of a scalar function and discuss its relation with the gradient operator. Lecture XVII Abstract We introduce te concept of directional derivative of a scalar function and discuss its relation wit te gradient operator. Directional derivative and gradient Te directional derivative

More information

Journal of Computational and Applied Mathematics

Journal of Computational and Applied Mathematics Journal of Computational and Applied Matematics 94 (6) 75 96 Contents lists available at ScienceDirect Journal of Computational and Applied Matematics journal omepage: www.elsevier.com/locate/cam Smootness-Increasing

More information

2.1 THE DEFINITION OF DERIVATIVE

2.1 THE DEFINITION OF DERIVATIVE 2.1 Te Derivative Contemporary Calculus 2.1 THE DEFINITION OF DERIVATIVE 1 Te grapical idea of a slope of a tangent line is very useful, but for some uses we need a more algebraic definition of te derivative

More information

Reflection Symmetries of q-bernoulli Polynomials

Reflection Symmetries of q-bernoulli Polynomials Journal of Nonlinear Matematical Pysics Volume 1, Supplement 1 005, 41 4 Birtday Issue Reflection Symmetries of q-bernoulli Polynomials Boris A KUPERSHMIDT Te University of Tennessee Space Institute Tullaoma,

More information

EFFICIENCY OF MODEL-ASSISTED REGRESSION ESTIMATORS IN SAMPLE SURVEYS

EFFICIENCY OF MODEL-ASSISTED REGRESSION ESTIMATORS IN SAMPLE SURVEYS Statistica Sinica 24 2014, 395-414 doi:ttp://dx.doi.org/10.5705/ss.2012.064 EFFICIENCY OF MODEL-ASSISTED REGRESSION ESTIMATORS IN SAMPLE SURVEYS Jun Sao 1,2 and Seng Wang 3 1 East Cina Normal University,

More information

Numerical Experiments Using MATLAB: Superconvergence of Nonconforming Finite Element Approximation for Second-Order Elliptic Problems

Numerical Experiments Using MATLAB: Superconvergence of Nonconforming Finite Element Approximation for Second-Order Elliptic Problems Applied Matematics, 06, 7, 74-8 ttp://wwwscirporg/journal/am ISSN Online: 5-7393 ISSN Print: 5-7385 Numerical Experiments Using MATLAB: Superconvergence of Nonconforming Finite Element Approximation for

More information

5 Ordinary Differential Equations: Finite Difference Methods for Boundary Problems

5 Ordinary Differential Equations: Finite Difference Methods for Boundary Problems 5 Ordinary Differential Equations: Finite Difference Metods for Boundary Problems Read sections 10.1, 10.2, 10.4 Review questions 10.1 10.4, 10.8 10.9, 10.13 5.1 Introduction In te previous capters we

More information

Copyright c 2008 Kevin Long

Copyright c 2008 Kevin Long Lecture 4 Numerical solution of initial value problems Te metods you ve learned so far ave obtained closed-form solutions to initial value problems. A closedform solution is an explicit algebriac formula

More information

The total error in numerical differentiation

The total error in numerical differentiation AMS 147 Computational Metods and Applications Lecture 08 Copyrigt by Hongyun Wang, UCSC Recap: Loss of accuracy due to numerical cancellation A B 3, 3 ~10 16 In calculating te difference between A and

More information

A MONTE CARLO ANALYSIS OF THE EFFECTS OF COVARIANCE ON PROPAGATED UNCERTAINTIES

A MONTE CARLO ANALYSIS OF THE EFFECTS OF COVARIANCE ON PROPAGATED UNCERTAINTIES A MONTE CARLO ANALYSIS OF THE EFFECTS OF COVARIANCE ON PROPAGATED UNCERTAINTIES Ronald Ainswort Hart Scientific, American Fork UT, USA ABSTRACT Reports of calibration typically provide total combined uncertainties

More information

4. The slope of the line 2x 7y = 8 is (a) 2/7 (b) 7/2 (c) 2 (d) 2/7 (e) None of these.

4. The slope of the line 2x 7y = 8 is (a) 2/7 (b) 7/2 (c) 2 (d) 2/7 (e) None of these. Mat 11. Test Form N Fall 016 Name. Instructions. Te first eleven problems are wort points eac. Te last six problems are wort 5 points eac. For te last six problems, you must use relevant metods of algebra

More information

Bootstrap confidence intervals in nonparametric regression without an additive model

Bootstrap confidence intervals in nonparametric regression without an additive model Bootstrap confidence intervals in nonparametric regression witout an additive model Dimitris N. Politis Abstract Te problem of confidence interval construction in nonparametric regression via te bootstrap

More information

Logistic Kernel Estimator and Bandwidth Selection. for Density Function

Logistic Kernel Estimator and Bandwidth Selection. for Density Function International Journal of Contemporary Matematical Sciences Vol. 13, 2018, no. 6, 279-286 HIKARI Ltd, www.m-ikari.com ttps://doi.org/10.12988/ijcms.2018.81133 Logistic Kernel Estimator and Bandwidt Selection

More information

Quantum Mechanics Chapter 1.5: An illustration using measurements of particle spin.

Quantum Mechanics Chapter 1.5: An illustration using measurements of particle spin. I Introduction. Quantum Mecanics Capter.5: An illustration using measurements of particle spin. Quantum mecanics is a teory of pysics tat as been very successful in explaining and predicting many pysical

More information

Mathematics 5 Worksheet 11 Geometry, Tangency, and the Derivative

Mathematics 5 Worksheet 11 Geometry, Tangency, and the Derivative Matematics 5 Workseet 11 Geometry, Tangency, and te Derivative Problem 1. Find te equation of a line wit slope m tat intersects te point (3, 9). Solution. Te equation for a line passing troug a point (x

More information

lecture 26: Richardson extrapolation

lecture 26: Richardson extrapolation 43 lecture 26: Ricardson extrapolation 35 Ricardson extrapolation, Romberg integration Trougout numerical analysis, one encounters procedures tat apply some simple approximation (eg, linear interpolation)

More information

Comment on Experimental observations of saltwater up-coning

Comment on Experimental observations of saltwater up-coning 1 Comment on Experimental observations of saltwater up-coning H. Zang 1,, D.A. Barry 2 and G.C. Hocking 3 1 Griffit Scool of Engineering, Griffit University, Gold Coast Campus, QLD 4222, Australia. Tel.:

More information

INTRODUCTION AND MATHEMATICAL CONCEPTS

INTRODUCTION AND MATHEMATICAL CONCEPTS INTODUCTION ND MTHEMTICL CONCEPTS PEVIEW Tis capter introduces you to te basic matematical tools for doing pysics. You will study units and converting between units, te trigonometric relationsips of sine,

More information

Chapter 5 FINITE DIFFERENCE METHOD (FDM)

Chapter 5 FINITE DIFFERENCE METHOD (FDM) MEE7 Computer Modeling Tecniques in Engineering Capter 5 FINITE DIFFERENCE METHOD (FDM) 5. Introduction to FDM Te finite difference tecniques are based upon approximations wic permit replacing differential

More information

Comonotonicity and Maximal Stop-Loss Premiums

Comonotonicity and Maximal Stop-Loss Premiums Comonotonicity and Maximal Stop-Loss Premiums Jan Dhaene Shaun Wang Virginia Young Marc J. Goovaerts November 8, 1999 Abstract In this paper, we investigate the relationship between comonotonicity and

More information

Analytic Functions. Differentiable Functions of a Complex Variable

Analytic Functions. Differentiable Functions of a Complex Variable Analytic Functions Differentiable Functions of a Complex Variable In tis capter, we sall generalize te ideas for polynomials power series of a complex variable we developed in te previous capter to general

More information

MAT244 - Ordinary Di erential Equations - Summer 2016 Assignment 2 Due: July 20, 2016

MAT244 - Ordinary Di erential Equations - Summer 2016 Assignment 2 Due: July 20, 2016 MAT244 - Ordinary Di erential Equations - Summer 206 Assignment 2 Due: July 20, 206 Full Name: Student #: Last First Indicate wic Tutorial Section you attend by filling in te appropriate circle: Tut 0

More information

REVIEW LAB ANSWER KEY

REVIEW LAB ANSWER KEY REVIEW LAB ANSWER KEY. Witout using SN, find te derivative of eac of te following (you do not need to simplify your answers): a. f x 3x 3 5x x 6 f x 3 3x 5 x 0 b. g x 4 x x x notice te trick ere! x x g

More information

Teaching Differentiation: A Rare Case for the Problem of the Slope of the Tangent Line

Teaching Differentiation: A Rare Case for the Problem of the Slope of the Tangent Line Teacing Differentiation: A Rare Case for te Problem of te Slope of te Tangent Line arxiv:1805.00343v1 [mat.ho] 29 Apr 2018 Roman Kvasov Department of Matematics University of Puerto Rico at Aguadilla Aguadilla,

More information

Department of Statistics & Operations Research, Aligarh Muslim University, Aligarh, India

Department of Statistics & Operations Research, Aligarh Muslim University, Aligarh, India Open Journal of Optimization, 04, 3, 68-78 Publised Online December 04 in SciRes. ttp://www.scirp.org/ournal/oop ttp://dx.doi.org/0.436/oop.04.34007 Compromise Allocation for Combined Ratio Estimates of

More information

HOMEWORK HELP 2 FOR MATH 151

HOMEWORK HELP 2 FOR MATH 151 HOMEWORK HELP 2 FOR MATH 151 Here we go; te second round of omework elp. If tere are oters you would like to see, let me know! 2.4, 43 and 44 At wat points are te functions f(x) and g(x) = xf(x)continuous,

More information

The derivative function

The derivative function Roberto s Notes on Differential Calculus Capter : Definition of derivative Section Te derivative function Wat you need to know already: f is at a point on its grap and ow to compute it. Wat te derivative

More information

5.1 We will begin this section with the definition of a rational expression. We

5.1 We will begin this section with the definition of a rational expression. We Basic Properties and Reducing to Lowest Terms 5.1 We will begin tis section wit te definition of a rational epression. We will ten state te two basic properties associated wit rational epressions and go

More information

Lecture 21. Numerical differentiation. f ( x+h) f ( x) h h

Lecture 21. Numerical differentiation. f ( x+h) f ( x) h h Lecture Numerical differentiation Introduction We can analytically calculate te derivative of any elementary function, so tere migt seem to be no motivation for calculating derivatives numerically. However

More information

New Fourth Order Quartic Spline Method for Solving Second Order Boundary Value Problems

New Fourth Order Quartic Spline Method for Solving Second Order Boundary Value Problems MATEMATIKA, 2015, Volume 31, Number 2, 149 157 c UTM Centre for Industrial Applied Matematics New Fourt Order Quartic Spline Metod for Solving Second Order Boundary Value Problems 1 Osama Ala yed, 2 Te

More information

Work and Energy. Introduction. Work. PHY energy - J. Hedberg

Work and Energy. Introduction. Work. PHY energy - J. Hedberg Work and Energy PHY 207 - energy - J. Hedberg - 2017 1. Introduction 2. Work 3. Kinetic Energy 4. Potential Energy 5. Conservation of Mecanical Energy 6. Ex: Te Loop te Loop 7. Conservative and Non-conservative

More information

The Verlet Algorithm for Molecular Dynamics Simulations

The Verlet Algorithm for Molecular Dynamics Simulations Cemistry 380.37 Fall 2015 Dr. Jean M. Standard November 9, 2015 Te Verlet Algoritm for Molecular Dynamics Simulations Equations of motion For a many-body system consisting of N particles, Newton's classical

More information

Logarithmic functions

Logarithmic functions Roberto s Notes on Differential Calculus Capter 5: Derivatives of transcendental functions Section Derivatives of Logaritmic functions Wat ou need to know alread: Definition of derivative and all basic

More information

ERROR BOUNDS FOR THE METHODS OF GLIMM, GODUNOV AND LEVEQUE BRADLEY J. LUCIER*

ERROR BOUNDS FOR THE METHODS OF GLIMM, GODUNOV AND LEVEQUE BRADLEY J. LUCIER* EO BOUNDS FO THE METHODS OF GLIMM, GODUNOV AND LEVEQUE BADLEY J. LUCIE* Abstract. Te expected error in L ) attimet for Glimm s sceme wen applied to a scalar conservation law is bounded by + 2 ) ) /2 T

More information

Applications of the van Trees inequality to non-parametric estimation.

Applications of the van Trees inequality to non-parametric estimation. Brno-06, Lecture 2, 16.05.06 D/Stat/Brno-06/2.tex www.mast.queensu.ca/ blevit/ Applications of te van Trees inequality to non-parametric estimation. Regular non-parametric problems. As an example of suc

More information

1. Consider the trigonometric function f(t) whose graph is shown below. Write down a possible formula for f(t).

1. Consider the trigonometric function f(t) whose graph is shown below. Write down a possible formula for f(t). . Consider te trigonometric function f(t) wose grap is sown below. Write down a possible formula for f(t). Tis function appears to be an odd, periodic function tat as been sifted upwards, so we will use

More information

EFFICIENT REPLICATION VARIANCE ESTIMATION FOR TWO-PHASE SAMPLING

EFFICIENT REPLICATION VARIANCE ESTIMATION FOR TWO-PHASE SAMPLING Statistica Sinica 13(2003), 641-653 EFFICIENT REPLICATION VARIANCE ESTIMATION FOR TWO-PHASE SAMPLING J. K. Kim and R. R. Sitter Hankuk University of Foreign Studies and Simon Fraser University Abstract:

More information

Continuity and Differentiability Worksheet

Continuity and Differentiability Worksheet Continuity and Differentiability Workseet (Be sure tat you can also do te grapical eercises from te tet- Tese were not included below! Typical problems are like problems -3, p. 6; -3, p. 7; 33-34, p. 7;

More information

FUNDAMENTAL ECONOMICS Vol. I - Walrasian and Non-Walrasian Microeconomics - Anjan Mukherji WALRASIAN AND NON-WALRASIAN MICROECONOMICS

FUNDAMENTAL ECONOMICS Vol. I - Walrasian and Non-Walrasian Microeconomics - Anjan Mukherji WALRASIAN AND NON-WALRASIAN MICROECONOMICS FUNDAMENTAL ECONOMICS Vol. I - Walrasian and Non-Walrasian Microeconomics - Anjan Mukerji WALRASIAN AND NON-WALRASIAN MICROECONOMICS Anjan Mukerji Center for Economic Studies and Planning, Jawaarlal Neru

More information

Combining functions: algebraic methods

Combining functions: algebraic methods Combining functions: algebraic metods Functions can be added, subtracted, multiplied, divided, and raised to a power, just like numbers or algebra expressions. If f(x) = x 2 and g(x) = x + 2, clearly f(x)

More information

SECTION 1.10: DIFFERENCE QUOTIENTS LEARNING OBJECTIVES

SECTION 1.10: DIFFERENCE QUOTIENTS LEARNING OBJECTIVES (Section.0: Difference Quotients).0. SECTION.0: DIFFERENCE QUOTIENTS LEARNING OBJECTIVES Define average rate of cange (and average velocity) algebraically and grapically. Be able to identify, construct,

More information

NONLINEAR SYSTEMS IDENTIFICATION USING THE VOLTERRA MODEL. Georgeta Budura

NONLINEAR SYSTEMS IDENTIFICATION USING THE VOLTERRA MODEL. Georgeta Budura NONLINEAR SYSTEMS IDENTIFICATION USING THE VOLTERRA MODEL Georgeta Budura Politenica University of Timisoara, Faculty of Electronics and Telecommunications, Comm. Dep., georgeta.budura@etc.utt.ro Abstract:

More information

One-Sided Position-Dependent Smoothness-Increasing Accuracy-Conserving (SIAC) Filtering Over Uniform and Non-uniform Meshes

One-Sided Position-Dependent Smoothness-Increasing Accuracy-Conserving (SIAC) Filtering Over Uniform and Non-uniform Meshes DOI 10.1007/s10915-014-9946-6 One-Sided Position-Dependent Smootness-Increasing Accuracy-Conserving (SIAC) Filtering Over Uniform and Non-uniform Meses JenniferK.Ryan Xiaozou Li Robert M. Kirby Kees Vuik

More information

University Mathematics 2

University Mathematics 2 University Matematics 2 1 Differentiability In tis section, we discuss te differentiability of functions. Definition 1.1 Differentiable function). Let f) be a function. We say tat f is differentiable at

More information

Discretization of Multipole Sources in a Finite Difference. Setting for Wave Propagation Problems

Discretization of Multipole Sources in a Finite Difference. Setting for Wave Propagation Problems Discretization of Multipole Sources in a Finite Difference Setting for Wave Propagation Problems Mario J. Bencomo 1 and William Symes 2 1 Institute for Computational and Experimental Researc in Matematics,

More information

Quasiperiodic phenomena in the Van der Pol - Mathieu equation

Quasiperiodic phenomena in the Van der Pol - Mathieu equation Quasiperiodic penomena in te Van der Pol - Matieu equation F. Veerman and F. Verulst Department of Matematics, Utrect University P.O. Box 80.010, 3508 TA Utrect Te Neterlands April 8, 009 Abstract Te Van

More information

Lecture 10: Carnot theorem

Lecture 10: Carnot theorem ecture 0: Carnot teorem Feb 7, 005 Equivalence of Kelvin and Clausius formulations ast time we learned tat te Second aw can be formulated in two ways. e Kelvin formulation: No process is possible wose

More information

Analysis of A Continuous Finite Element Method for H(curl, div)-elliptic Interface Problem

Analysis of A Continuous Finite Element Method for H(curl, div)-elliptic Interface Problem Analysis of A Continuous inite Element Metod for Hcurl, div)-elliptic Interface Problem Huoyuan Duan, Ping Lin, and Roger C. E. Tan Abstract In tis paper, we develop a continuous finite element metod for

More information

INTRODUCTION AND MATHEMATICAL CONCEPTS

INTRODUCTION AND MATHEMATICAL CONCEPTS Capter 1 INTRODUCTION ND MTHEMTICL CONCEPTS PREVIEW Tis capter introduces you to te basic matematical tools for doing pysics. You will study units and converting between units, te trigonometric relationsips

More information

Functions of the Complex Variable z

Functions of the Complex Variable z Capter 2 Functions of te Complex Variable z Introduction We wis to examine te notion of a function of z were z is a complex variable. To be sure, a complex variable can be viewed as noting but a pair of

More information

HOW TO DEAL WITH FFT SAMPLING INFLUENCES ON ADEV CALCULATIONS

HOW TO DEAL WITH FFT SAMPLING INFLUENCES ON ADEV CALCULATIONS HOW TO DEAL WITH FFT SAMPLING INFLUENCES ON ADEV CALCULATIONS Po-Ceng Cang National Standard Time & Frequency Lab., TL, Taiwan 1, Lane 551, Min-Tsu Road, Sec. 5, Yang-Mei, Taoyuan, Taiwan 36 Tel: 886 3

More information

Calculation of Bayes Premium for Conditional Elliptical Risks

Calculation of Bayes Premium for Conditional Elliptical Risks 1 Calculation of Bayes Premium for Conditional Elliptical Risks Alfred Kume 1 and Enkelejd Hashorva University of Kent & University of Lausanne February 1, 13 Abstract: In this paper we discuss the calculation

More information

3.4 Worksheet: Proof of the Chain Rule NAME

3.4 Worksheet: Proof of the Chain Rule NAME Mat 1170 3.4 Workseet: Proof of te Cain Rule NAME Te Cain Rule So far we are able to differentiate all types of functions. For example: polynomials, rational, root, and trigonometric functions. We are

More information

Basic Nonparametric Estimation Spring 2002

Basic Nonparametric Estimation Spring 2002 Basic Nonparametric Estimation Spring 2002 Te following topics are covered today: Basic Nonparametric Regression. Tere are four books tat you can find reference: Silverman986, Wand and Jones995, Hardle990,

More information

Quaternion Dynamics, Part 1 Functions, Derivatives, and Integrals. Gary D. Simpson. rev 01 Aug 08, 2016.

Quaternion Dynamics, Part 1 Functions, Derivatives, and Integrals. Gary D. Simpson. rev 01 Aug 08, 2016. Quaternion Dynamics, Part 1 Functions, Derivatives, and Integrals Gary D. Simpson gsim1887@aol.com rev 1 Aug 8, 216 Summary Definitions are presented for "quaternion functions" of a quaternion. Polynomial

More information

More on generalized inverses of partitioned matrices with Banachiewicz-Schur forms

More on generalized inverses of partitioned matrices with Banachiewicz-Schur forms More on generalized inverses of partitioned matrices wit anaciewicz-scur forms Yongge Tian a,, Yosio Takane b a Cina Economics and Management cademy, Central University of Finance and Economics, eijing,

More information

Pre-Calculus Review Preemptive Strike

Pre-Calculus Review Preemptive Strike Pre-Calculus Review Preemptive Strike Attaced are some notes and one assignment wit tree parts. Tese are due on te day tat we start te pre-calculus review. I strongly suggest reading troug te notes torougly

More information

Quantum Numbers and Rules

Quantum Numbers and Rules OpenStax-CNX module: m42614 1 Quantum Numbers and Rules OpenStax College Tis work is produced by OpenStax-CNX and licensed under te Creative Commons Attribution License 3.0 Abstract Dene quantum number.

More information

Chapter 2. Limits and Continuity 16( ) 16( 9) = = 001. Section 2.1 Rates of Change and Limits (pp ) Quick Review 2.1

Chapter 2. Limits and Continuity 16( ) 16( 9) = = 001. Section 2.1 Rates of Change and Limits (pp ) Quick Review 2.1 Capter Limits and Continuity Section. Rates of Cange and Limits (pp. 969) Quick Review..... f ( ) ( ) ( ) 0 ( ) f ( ) f ( ) sin π sin π 0 f ( ). < < < 6. < c c < < c 7. < < < < < 8. 9. 0. c < d d < c

More information

Order of Accuracy. ũ h u Ch p, (1)

Order of Accuracy. ũ h u Ch p, (1) Order of Accuracy 1 Terminology We consider a numerical approximation of an exact value u. Te approximation depends on a small parameter, wic can be for instance te grid size or time step in a numerical

More information

3. THE EXCHANGE ECONOMY

3. THE EXCHANGE ECONOMY Essential Microeconomics -1-3. THE EXCHNGE ECONOMY Pareto efficient allocations 2 Edgewort box analysis 5 Market clearing prices 13 Walrasian Equilibrium 16 Equilibrium and Efficiency 22 First welfare

More information

Bandwidth Selection in Nonparametric Kernel Testing

Bandwidth Selection in Nonparametric Kernel Testing Te University of Adelaide Scool of Economics Researc Paper No. 2009-0 January 2009 Bandwidt Selection in Nonparametric ernel Testing Jiti Gao and Irene Gijbels Bandwidt Selection in Nonparametric ernel

More information

Derivatives of Exponentials

Derivatives of Exponentials mat 0 more on derivatives: day 0 Derivatives of Eponentials Recall tat DEFINITION... An eponential function as te form f () =a, were te base is a real number a > 0. Te domain of an eponential function

More information

Chapter 1. Density Estimation

Chapter 1. Density Estimation Capter 1 Density Estimation Let X 1, X,..., X n be observations from a density f X x. Te aim is to use only tis data to obtain an estimate ˆf X x of f X x. Properties of f f X x x, Parametric metods f

More information

2.11 That s So Derivative

2.11 That s So Derivative 2.11 Tat s So Derivative Introduction to Differential Calculus Just as one defines instantaneous velocity in terms of average velocity, we now define te instantaneous rate of cange of a function at a point

More information

Complexity of Decoding Positive-Rate Primitive Reed-Solomon Codes

Complexity of Decoding Positive-Rate Primitive Reed-Solomon Codes 1 Complexity of Decoding Positive-Rate Primitive Reed-Solomon Codes Qi Ceng and Daqing Wan Abstract It as been proved tat te maximum likeliood decoding problem of Reed-Solomon codes is NP-ard. However,

More information