Cohorts are the aggregate of individuals who experience the same event within the same time interval. Cohorts can be based on people born in a given year, for example in 1940 or within a span of years, e.g. born in 1940-1944. The year of birth is here the defining event for cohorts. The health differs between cohorts. This article focuses on the protective and detrimental cohort effect in relation to the risk of death from apoplexy (stroke). A dummy variable method is recommended to describe the changing cohort effect over a century. Likewise it is shown how information in the data diagonals in an age-period-cohort model can be transferred to a time series model.