LARGE SAMPLE TESTS AND CONFIDENCE INTERVALS FOR MORTALITY RATES
Abstract
In computing mortality rates from insurance data, the unit of measurement used is frequently based on number of policies or amount of insurance rather than on lives. Then the death of one person may result in several units of 'death' with respect to the investigation; moreover, the number of units per individual may vary noticeably. Thus the usual large sample methods of obtaining significance tests and confidence intervals for the true value of the mortality rate are not applicable to these situations. If the number of units associated with each person in the investigation were known, accurate large sample results could be obtained; however, determination of the number of units associated with each individual would require an extremely large amount of work. The article presents some valid large sample tests and confidence intervals for the mortality rate which do not require much work and are reasonably efficient. More general situations are also considered.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 24, 1949
- Accession Number
- AD0603828
Entities
People
- John E. Walsh
Organizations
- RAND Corporation