Period life tables from longitudinal data (cont.)

Once all records in the specified period are split – so you will have considerably more records than before, but each is no longer than 1 person-year – you can ask the package to collapse (aggregate) them, summing the person-years and the events for each single year of age.

It is then a simple matter to produce age-specific mortality rates – these will be events at age x divided by summed person-years at age x.

We now have a set of age-specific central rates. These are _nm_x rates.

A life table can now be formed in exactly the way demonstrated in the Life Tables II session of Module 1. If done manually (by spreadsheet) this will involve the use of Chiang’s formula to convert _nm_x rates to _nq_x risks – the starting points for life table construction. Censoring has already been accounted for in the person-years derivation.

However, having reached this stage using a statistical package it is more common to allow the package to continue with survival analysis routines to form a life table. This has advantages of being able to test between sub-groups in the data, produce graphs etc. It is also the basis for going further and modelling in multivariable statistical models. The models of choice here would be Poisson models.