Statistical inference for familial disease models assuming exchangeability


Binary data often occur in epidemiology studies clustered by families. When an assumption of exchangeability of response is reasonable, the exchangeable multinomial distribution may be used to model binary and multinomial responses. Conditional probabilities of a new family member being diagnosed given family members previously diagnosed are derived. Applications are made to real familial disease data on IPF and ALS.

Publication Title

Statistics and Probability Letters