Pooled Attributable Risk for 2 x 2 x K Tables Using Weighted Least Squares Estimation

The attributable risk is the fraction of outcomes in the population that would be reduced if the probability for a (higher probability) group were reduced to the probability for another (lower probability) group (Fleiss, et. al). We will use methods of weighted least squares (Koch, et. al) and maximum likelihood (Dobson and Barnett) to compute odds ratios and the pooled odds ratio, and weighted least squares to compute attributable risks and the pooled attributable risk, which we introduce here using a Mantel- Haenszel- like definition. Although odds ratios and attributable risks can be calculated using specialized statistical software, we will show how these can be obtained using Microsoft. Excel array formulas and V B A programming.