Introduction

A previous module and session (Module 1, Session 7) dealt with the construction of life tables up to the lx column, which for many purposes is quite adequate.

We stated there that life tables are formed from one of two sources of data: either from a given set of failure rates, or from individual level data such as that collected in a survey or longitudinal study.

For a mortality life table the failure rates would be age-specific death rates and these may have been produced from vital registration and census data. If the rates are derived from data collected in a particular year, or over a small number of years, the table will be a period table, that is referring to a particular period of time, not a cohort of people. It is however quite possible, although less common, to form rates from data derived from following a group of people through time and a life table produced from these rates would be a cohort life table.

Using the rates to form a life table is relatively simple – a first step being to convert the rates (which will be central rates – _nm_x rates) into risks or probabilities of dying (_nq_x rates) which is the starting point for all life tables.

This session deals with data derived from individual-level data from a survey or longitudinal study, rather than from aggregated data. It requires rather more processing to form the necessary _nq_x rates but once formed the life table proceeds as already described.

Complications arise in the form of censoring and this session deals with that process. Period or cohort tables can be produced. Because the processing of such data is rather more complicated, and in greater volume, it invariably involves the use of a computer and this has advantages in that a confidence interval of the l_x column can be constructed and this facilitates statistical testing between the survival columns of different groups – which is effectively survival analysis.