35

6. Calibration method The following three steps are taken separately for the two sexes  :  We take the exposures  ,   and    with                        and observed deaths   for the relevant Western European countries, with      and    . This concerns in all cases the sum of all exposures and the sum of all deaths in the respective countries, including the Netherlands. The parameters   were then determined in such a way that the Poisson likelihood function for the observed deaths is as large as possible for the given exposures:         requiring the sum of the elements of  elements of  . To obtain a unique specification of the three vectors we normalise by  over   to be equal to  and the sum of the  over   to be equal to .  Data after 2014 are not available for all the relevant countries. For this reason, the values of  then applied        en  , by means of    with                     The maximum likelihood method is now applied to the data for the Netherlands to determine             as previously. Once again, normalisation takes place by requiring the sum of the elements in   over  and  In a fourth and final step, the four time series are used, namely       ,    (now including the year 2015) and  over   to be respectively  and .     the parameters  and matrix . On the assumption that the variables     are independently and identically distributed and have a four-dimensional normal distribution with average (0,0,0,0) and covariance matrix , we select the estimators for  and  in such a way6 that the likelihood of these time series is maximised. 7. Simulation of the time series In order to be able to simulate the scenarios for the timeseries         to estimate            in the previous step are determined up to and including 2014. Linear extrapolation is     , samples from a normal distribution with an average of (0,0,0,0) and covariance matrix C must be generated. This can be done by multiplying a (row) vector  with four independently standard normally distributed variables by a matrix H, which satisfies the condition  , in other words by means of  . In the list of parameters in the publication and the accompanying Excel spreadsheet, a Cholesky H matrix has therefore also been included in addition to the covariance matrix C. 6 The working group made use of the R package, systemfit, for this with the options method=”SUR” and methodResidCov=”noDfCor”. Projection Table AG2016 Appendix A 35

36 Online Touch Home


You need flash player to view this online publication