580 – Disease Prediction
Inference Using Impact Numbers for a Multinomial Sampling Design
Khairul Islam
Eastern Michigan University
Tanweer J. Shapla
Eastern Michigan University
Impact numbers reflect the number of people specific to a population among whom one outcome or case is attributable to the exposure of the risk factor. When being exposed to the risk factor is significant for the development of the disease as measured by the standard effect measures such as relative risk, risk difference or attributable risk as appropriate, the impact numbers provide very useful information not possible otherwise. To date, a few studies exist in literature that takes impact numbers into account in making inference. In particular, while confidence interval estimates of impact numbers are investigated for cohort and case-control studies, they are not yet documented adequately for a multinomial sampling design. This paper provides confidence interval estimates of impact numbers for a multinomial sampling design. Real life example and simulation studies are considered to justify performance of these methods.