For mortality forecasting using APC model (Renshaw and Haberman, 2005 or Currie, 2006) , we first estimate the 3 compenents: Age, time and Cohort effect. then, we forcast the time and Cohort effect in the future. For Time trend, we use ARMA or ARIMA models. The cohort trend is generally changing, so, we use only the recent observed trend.
Q1: how to select this time range ?
Q2: which forecasting technics to use?
look at the favourite links page of my website and look at the electronic stats book Deborah Hilton Statistics Online
http://sites.google.com/site/deborahhilton/
A slightly different approach for forecasting mortality is the functional data approach developed by Rob Hyndman and his colleagues. One of the advantages of this approach is that it has an inbuilt capacity to determine the best time horizon to be used for forecasting, which you may be able to adapt to your APC forecasting model. If this is of interest to you, you may like to check one of their recent publications in Demography--the journal published by the American Population Association. "Coherent mortality forecasting: the product-ratio method with functional time series models." Demography. 50(1):261-83. 2013.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics