APPENDIX A DERIVAION OF JOIN FAILRE DENSIIES I his Appedi we prese he derivaio o he eample ailre models as show i Chaper 3. Assme ha he ime ad se o ailre are relaed by he cio g ad he sochasic are o his relaioship ca be represeed by reaig oe or wo o he parameers o g as radom variables. We cosider he ollowig or eample orms: i g ii g iii g ad iv g e e. I each case we irodce radomess io he cio by reaig he parameer as a radom variable havig disribio. sig ad he rasormaio o variables we cosrc he margial probabiliy disribio o sage. ha is d g. A. d Oce he margial disribio o sage is obaied we he cosrc he joi ailre desiy sig he codiioig relaio i.e.. A. 7
73 I Eq. A. he codiioal desiy is obaied by sig he well-kow relaioship bewee a desiy ad is hazard cio: { } { } g g d z z d z z ep ep. A.3 We assme ha he codiioal bivariae hazard cio o age give sage may be saed as: z A.4 so ha he deiiios o he cios ad g deermie he codiioal hazard ad limaely he bivariae lie disribio. I here i order o ocs o he cios g we assme ha ad are simple liear cios i.e. ad. hs we assme z. A.5 der his modelig orma we may obai he bivariae lie disribios correspodig o orms i ii iii ad iv respecively as ollows: ep A.6 Eq. 3.7 6 3 3 ep A.7 Eq. 3.8
74 ep A.8 Eq. 3.9 ad ep. A.9 Eq. 3. A. Derivaio o Eq. A.6 For case i g solvig or yields: ad d d so:. A. he codiioal bivariae hazard cio o age give sage ca be cosrced as: z A. Sbsiig Eq. A. io Eq.A.3 we obai d ep
75 ep A. Sbsiig Eqs. A. ad A. io Eq. A. we derive ep. A. Derivaio o Eq. A.7 For case ii g solvig or yields: ad d d so:. A.3 he codiioal bivariae hazard cio o age give sage ca be cosrced as: z A.4 Sbsiig Eq. A.4 io Eq.A.3 we obai
76 d ep 6 3 3 ep A.5 Sbsiig Eqs. A.3 ad A.5 io Eq. A. we derive 6 3 3 ep. A.3 Derivaio o Eq. 3.9 For case ii g solvig or yields: ad d d so:. A.6 he codiioal bivariae hazard cio o age give sage ca be cosrced as: z A.7 Sbsiig Eq. A.7 io Eq.A.3 we obai
77 d ep ep A.8 Sbsiig Eqs. A.6 ad A.8 io Eq. A. we derive ep. A.4 Derivaio o Eq. 3. For case iv g e e solvig or yields: ad d d so:. A.9 he codiioal bivariae hazard cio o age give sage ca be cosrced as: z ep ep. A. Sbsiig Eq. A. io Eq.A.3 we obai
78 d ep ep ep d ep Le he d ep ep ep ep A. Sbsiig Eqs. A. ad A.9 io Eq. A. we derive
79 ep.
APPENDIX B HE BIVARIAE DIRAC DELA FNCION I his Appedi we irodce he bivariae Dirac Dela cio ha is sed i Chaper 6 o model he logeviy o he sigle-i sysem i he PM cycle o he ARPM model. We begi wih he discssio o he ivariae Dirac Dela cio ad is properies. he we eed he ivariae cio o wo dimesios ad prese he resls or he bivarie Dirac Dela cio. B. ivariae Dirac Dela Fcio Cosider he ollowig cio: ε ε ε > F ε B. oherwise. Figre B.. shows he plo o F ε. Geomerically as ε we have he ollowig iegral: F ε /ε. he F ε d. B. Le δ be he limiig cio o F ε as ε. δ is called ivariae Dirac Dela or i Implse cio or F ε see Spiegel [965]. 8
F ε /ε ε Figre B.. Plo o F ε. We desigae he ivariae Dirac Dela cio as δ ha has he ollowig properies: δ d B.3 i δ or ay coios cio φ. B.4 ii φ d φ Noe ha δ has a siglariy o iiie vale a ad is eqal o zero a all oher vales o. Propery i gives he Dirac Dela cio he characerisic o a probabiliy desiy cio. Propery ii helps i akig Laplace rasorm o he Dirac Dela cio. he Laplace rasorm o δ is give by 8
L s v { } δ s s δ e d e B.5 B. Bivariae Dirac Dela Fcio We ow cosrc he bivariae Dirac Dela cio wih aalogical properies as i ivariae case. Cosider he ollowig cio: ε ε ε ε > F ε B.6 oherwise. Figre B.. shows he plo o F ε. Geomerically as ε F ε /ε. he we have he ollowig iegral: F ε dd. B.7 Le δ be he limiig cio o as ε. δ is called bivariae Dirac F ε Dela or i Implse cio or. We desigae he ivariae Dirac Dela cio as δ ha has he ollowig properies: F ε he bivariae Dirac Dela cio or his PM cycle δ ollowig properies: has he i δ dd B.8 ii δ φ dd φ or ay coios cio φ. B.9 8
F ε /ε ε /ε ε Figre B.. Plo o F ε. Propery i gives he bivariae Dirac Dela cio he characerisic o a bivariae probabiliy desiy cio. Propery ii helps i akig Laplace rasorm o he bivariae Dirac Dela cio. he Laplace rasorm o δ is L { } δ s v s v s v δ e dd e. B. 83