Page 123 - Contributed Paper Session (CPS) - Volume 3
P. 123

CPS1965 Chin Tsung R. et al.
            (1992)  estimated  the  parameters  of  (1)  using  a  two-stage  method  with
                                           ̂
                          ̂
            restrictions ∑  () = 0 and ∑  () = 1  to ensure a unique solution for the
                         
                                         
            system of equations of the model. The singular value decomposition (SVD)
            approach was applied to the matrix of centered age profiles ( ) − ̂(),
                                                                            ,
            which allows a first estimation of parameters () and (). A second step
                                                                    ̂
                                                          ̂
            based on the refitting of () on the number of deaths is usually suggested to
                                     ̂
            assure a better convergence between the estimated and observed deaths. The
            aim is to find the () such that:
                              ̂

                                          
                                                                        ̂
                                                                    ̂
                         ∑      (, ) = ∑  (, ) (̂() + () ()),             (2)
                            = 1         = 1

            where (, ) is  the  number  of  deaths  of  age  in  time,  and (, ) is  the
            exposure to risk of age  in time . For this method, () and () are fixed.
            The  adjusted  ()  is  then  extrapolated  using  autoregressive  integrated
                           ̂
            moving average (ARIMA) models. The LC model uses a random walk with drift
            model, which can be expressed as:

                                    () =  (  −  1) +    +  (),                                       (3)

            where  is known as the drift parameter and measures the average annual
            change in the series, and () is an uncorrelated error.
                Zhao (2012) introduced a modified approach to LC model for analysing
            short base period age-specific data using linearized cubic splines and other
            additive functions to estimate the  unknown functions, () and (). Zhao
            observed  that  predicted  mortality  curves  were  not  smooth  and  fluctuate
            significantly over the age range. In this study, we applied the modified model
            by  Zhao  (2012)  for  the  Malaysian  mortality  data  (Department  of  Statistics
            Malaysia) for the period 1991-2015 for males and females.

            2.   Methodology
                According to Zhao  (2012),  a  piecewise cubic spline function with knots
             = { ,  , … ,  } can be expressed as
                              
              
                        2
                     1

                     () =  ()  +  ()    +∙∙∙ + ()                    (4)
                          0    0 ≤≤ 1  1   1 ≤≤ 2      ≤≤ +1

                                              2
                                   3
             where () = ( − ) + ( − ) + ( − ) + ,  = 0,1, …  and 
                                                                                 ≤≤ +1
            is the indicator function with value 1 on the interval [, +1] and 0 elsewhere.
            It is assumed that  () and its first and second derivatives are continuous
                                
            and  therefore  we  have  −1()  =  (),   −1()  =  (),   −1()  =   ()  for
                                                                                ′′
                                                                      ′′
                                                     ′
                                                               ′
             = 1,2,… , . Hence   () can be expressed as:
                                  

                                                               112 | I S I   W S C   2 0 1 9
   118   119   120   121   122   123   124   125   126   127   128