The Comparison of Multidimensional Rate Tables - A Simulation Study.
HARVARD UNIV CAMBRIDGE MASS DEPT OF STATISTICS
Pagination or Media Count:
The results of a Monte Carlo study of the problem of estimating the ratio R of rates from two multidimensional rate tables are presented. Using large simulated tables, four different methods are compared for their ability to estimate the ratio when R is constant from cell to cell. The ratio of direct standardized rates used on both the raw data and tables first fitted by log linear models gives unbiased estimates with a slight increase in precision introducted by first fitting the tables. However, the ratio of indirect standardized rates and Smear and Sweep followed by standardizing gives badly biased estimates in some cases. Author
- Statistics and Probability