Analysis of an in vivo assay for inactivated polio vaccine

Approaches to the statistical analysis of data from the rat bioassay for Inactivated Polio Vaccine are discussed and compared, using data from a recent collaborative study The measured response, an antibody titre obtained from a doubling dilution series, did not satisfy the requirements of normality and homogeneity of variance necessary for the standard parallel line model for quantitative data. Laboratory-specific cut-offs can be defined, to reduce the data to "positive" or "negative" responses, allowing valid analysis with the probit method.