For the sake of example, if we assume that the calender year of observation is 1st Jan 2005 to 31st Jan 2005, and current date is 1st Jan 2006, someone born in 2004 and married sometime in 2005 (thus aged x on a calender year basis) could be aged either 0 (x-1) or 1 (x) 'last birthday' (assuming uniform distribution).
Then, shouldn't Px*(t) = 1/2 * [Px-1(t) + Px(t)]?
Last edited: Mar 27, 2011