This paper by David R. Cox discusses the analysis of censored failure times using regression models and life-tables. It assumes that each individual has values of one or more explanatory variables, and the hazard function is a function of these variables and unknown regression coefficients multiplied by an arbitrary and unknown function of time. A conditional likelihood is used to make inferences about the unknown regression coefficients. Life tables are an important statistical technique used in medical statistics and actuarial science. However, their formal statistical theory has not been widely explored. Kaplan and Meier (1958) provided a comprehensive review of earlier work, and Chiang (1968) explored the connection with birth-death processes. This paper extends the results of Kaplan and Meier to the comparison of life tables and the incorporation of regression-like arguments into life-table analysis. The paper considers a population of individuals, where for each individual, either the time to failure or the time to loss or censoring is observed. The survivor function, $ \mathcal{F}(t) $, is defined as the probability that the failure time is greater than t. The hazard function, $ \lambda(t) $, is the age-specific failure rate. The product law of probability gives the survivor function as a product integral, which can be expressed as an exponential function if the hazard function is integrable. The paper also discusses the product-limit method, which is used to estimate the survivor function when there is censoring. The method is closely related to procedures for combining contingency tables and has applications in industrial reliability studies and medical statistics. The paper outlines the generalization of these methods and their relevance to practical situations where sampling fluctuations are significant.This paper by David R. Cox discusses the analysis of censored failure times using regression models and life-tables. It assumes that each individual has values of one or more explanatory variables, and the hazard function is a function of these variables and unknown regression coefficients multiplied by an arbitrary and unknown function of time. A conditional likelihood is used to make inferences about the unknown regression coefficients. Life tables are an important statistical technique used in medical statistics and actuarial science. However, their formal statistical theory has not been widely explored. Kaplan and Meier (1958) provided a comprehensive review of earlier work, and Chiang (1968) explored the connection with birth-death processes. This paper extends the results of Kaplan and Meier to the comparison of life tables and the incorporation of regression-like arguments into life-table analysis. The paper considers a population of individuals, where for each individual, either the time to failure or the time to loss or censoring is observed. The survivor function, $ \mathcal{F}(t) $, is defined as the probability that the failure time is greater than t. The hazard function, $ \lambda(t) $, is the age-specific failure rate. The product law of probability gives the survivor function as a product integral, which can be expressed as an exponential function if the hazard function is integrable. The paper also discusses the product-limit method, which is used to estimate the survivor function when there is censoring. The method is closely related to procedures for combining contingency tables and has applications in industrial reliability studies and medical statistics. The paper outlines the generalization of these methods and their relevance to practical situations where sampling fluctuations are significant.