MATH4/68181: Extreme values and financial risk Semester 1 Solutions to problem sheet for Week 2 ‘The density function of the Fréchet density function of the Weibull extreme value distribution is a(—2) The mean of the Gumbel extreme value dis [00 — exp(—2)} dx ‘The density function of the Gumbel extreme value distribution is exp(—z) exp{—exp(—2)} ‘onl me value distribution is az~’ ond exp{-27*} exp{—(-=)*}, istribution is = [nye vay [y= exr(-2) @ I” aoe {—y)d -- gf vernal a = -Are+1) = Tq) ‘The mean of the Fréchet extreme value distribution is [x02 Lexp (2 *}de afr exp {-2 *} dz a[>e exp {-2 *}dz a [vet hdr [y= = dr of vex (-v} dy dy aw a [vex ( v} - [vem ay ‘The mean of the Weibull extreme value distribution is ° [car exp {—(-2)"} dz 0 - wo f(a) exp (-(-2)"} de = ef’ yet vyde [y= (-2)% 0 dx = of yesp {vayaly [oon = [ive Cap at tay = Putem-nay = =f ve euay = -r(2+1) [Pee -a)em-em(-o)}de = [dryer updy (y= ex(-2)] 3. Since - Sfvrecay la~0 a = Sre+y oo - "a, the variance of the Gumbel extreme value distribution is I” (1) — [I"(1)}. Since [ Pax exp(-2-}de = f exp (-y}dy [y= exp(—2)] = PU-2/a), the vatiance of the Fréchet extreme value distribution is T(1 ~ 2/a) = T?(1 = 1/a). Since 0 [. wa(—a)*"* exp {—(-2)*}de = f vl exp{-u}dy [y= (-z)"] = T/a+1), the variance of the Weibull extreme value distribution is I (2/a-+ 1) ~ 1 (I/a +1). . We have A(x) = A (ant + 8,) <> exp {-nexp(—2)} = exp {— exp (ant ~ 6n)} > nexp(—z) = exp(—anr — Bn) 4 logn — 2 = —anr— Bp, Hence, the result We have BY(2) = Gq (ane + Bn) 4 exp (-ne9) = exp (— (an + Bn)*) ne = (ane + Bp)? ee NE = an t+ Bn Hence, the result6. We have W(x) = Wa (ane + Bn) exp (-n(=2)*) = exp (— (002 ~ f)") e ant — By)® ° an + Br Hence, the result 7. Note that w(F) = co and take 9(¢) = 1. Then im Lae) _ exp(-t=2) _ ae 1 F® expt) PC) So, the exponential cdf F(z) = 1—exp(—z) belongs to the Gumbel domain of attraction, 8. Note that w(F) = oo and take 4(t) = 1. Then =F(t+2) _)1-U-exsp(-t- 2)" _ 4 wep(-t-2) org Bao en ar 7 Bena tes = exp(-2) So, the exponentiated exponential edf F(x) = [1 -exp(—z)]* belongs to the Gumbel domain of attraction Note that w(F) = 1. Then So, the uniform|0, 1] edf F(x) =r belongs to the Weibull domain of attraction, 10. We have F(t) _ (Kj(tx))* be FO > TR" So, the Pareto cdf F(x) = 1 — (K/x)* belongs to the Fréchet domain of attraction. 11, Firstly, suppose that G belongs to the max domain of attraction of the Gumbel extreme value distribution. Then, there must exist a strictly positive function, say h(t), such that im LEGU+ zhi) erste) 1 G® = exp(-2) for every x € (—00, 00). But, using L’Hopital’s rule, we note that fm LEA) [Leah f+ 2h) conte) 1 FO) Sate) ® _ [r+ 2h @] + 2h) pa(es an(yye? ~ 8 as a a «EERO exp {oG(t) — 0G (t+ xh(t))} 3bese] oe+ eno 2h O) g (t+ 2h(Q) pyp-t ~ By li ] 1-G(t+ ah(t)) Re t ew {e~¢ 1 [r=] ot +2K(e)) = he ae) G (t+ xh(t))o? (Aa , G (t+ zh(t)) ~ ute) 1 Oh) G(t+ah(t)) Poo *| CO _ [: =G(t+ enon] ~ —6W lexp(-z)]" = exp(-bz) for every x € (—00, 00). So, it follows that F also belongs to the max domain of attraction of the Gumbel extreme value distribution with, Tita, Pr {an (My = bp) <2} for some suitable norming constants ay > 0 and by, Secondly, suppose that G belongs to the max domain of attraction of the Fréchet extreme value distribution. Then, there must exist a J <0 such that for every x > 0. But, using L'Hopital’s rule, we note that 1-Fliz) _ y,, 2flt2) ATO 7 ARG '») fos))""* 11 Glee)? = tim 22 a | eS | exp {eG(!) ~ eG (tz)} sates) fata T= GW) _ = Gtx) f1- Gtx] - Inara [sel - af[ Sey 2 for every a > 0. So, it follows that F also belongs to the max domain of attraction of the Fréchet extreme value distribution with stim, Pr {ay (My — ba) < exp (2)12, for some suitable norming constants a, > 0 and by. Thirdly, suppose that G belongs to the max domain of attraction of the Weibull extreme value distribution. Then, there must exist a a > 0 such that 1-G(w(G) tz) POT Gwe)” for every x > 0, But, using L’Hopital’s rule, we note that vim 2000 Gw(F Oa gw) Gwe) x exp {cG(w(F) — #) — eG (w(#) — tx)} Jam 220000) = tx) [1 ~ Gw(F) = tx) |? ~ 88 g(wt [Saeea tim Lo G(w(F) = te) [1 Gw(P) tx) ~ A T=Gwir) 9 Faces] 1 = Gw(F) = tx)]" ~ pa ee ae So, it follows that F also belongs to the max domain of attraction of the Weibull extreme =) value distribution with tiga, Pr {ay (My ~ by) <2} = exp for some suitable norming constants ay, > 0 and by, First, suppose that G belongs to the max domain of attraction of the Gumbel extreme value distribution. Then, there must exist a strictly positive function, say A(t), such that Gis ah(o)) 1-GQ lim, sou exp(-2) for every x € (—00, 00). But, using I'Hopital’s rule, we note that Fx (t+ 2h(0) (14 2h'(0) fx (C+ 2h) touthy 1 Pet) = i, 710) 1 (14 2h") 9+ 2h) para any]? -_; eax cme - 1-U-sa@w) yy «| 1G) ] lasnee sor] (+20) 9+ 2h) raid zo)" 1 = lim, OI ute) at) T = exp(-br)for every 2 € (0,00). So, it follows that F also belongs to the max domain of attraction of the Gumbel extreme value distribution with stim, Pr {ay (My ~ by) <2} = exp {—exp(~b2)} for some suitable norming constants ay, > 0 and by, Second, suppose that G belongs to the max domain of attraction of the Fréchet extreme value distribution. Then, there must exist ac <0, such that Gitte) $2 Tem for every > 0. But, ing !'Hopital’s rule, we note that B31 Fe@® Gey] my ao 1-G(tz))>" 1-6) Glitz) G(tz))" 2240. [He] 1a for every + > 0. So, it follows that F also belongs to the max domain of attraction of the Fréchet extreme value distribution with Tim Pr {an (My by) < 2} = exp ( a) for some suitable norming constants a,, > 0 and 6, Third, suppose that G belongs to the max domain of attraction of the Weibull extreme value distribution. Then, there must exist ac > 0, such that 23 Gq) — 9) for every x <0. But, using 'Hopital’s rule and results in Section 2.2, we note that tml Fw)=m) _ y 23 T— Fw(F) 5 fa] tian 22(W(F) = tx) (Cw) ~ tx))* * Ga gw) — 4) LOG) 8) 1-(1— 8)G(w(F) - 4) 7"? T- (1 8)G(w(F) — ta) mn 22 (UCR) = tz) JL Ea gwtF) 2) G(w(F) — tx))* 3(w(F) —) GwtF) 6= Im wt) =i G(w(P) oT G(R) (TGF) — 9) G(u(F) = Im at = ab for every x < 0. So, it follows that F also belongs to the max domain of attraction of the Weibull extreme value distribution with slim, Pr {an (My ~ bu) <2} = exp {-(-2)**} for some suitable normaing constants a, > 0 and by 13. We have Hence, the result 14. We have Hence, the result