On the age-period-cohort analysis of suicide rates.

Cohort analysis has become a popular method of examining national trends in suicide rates. Most of the studies investigating this phenomenon have reported a cohort effect to be present. Using a graphical approach, this paper places cohort analysis within the broader framework of age-period-cohort analysis. It is shown that published reports may have failed to identify cohort effects due to using only portions of the available data. With a simple mathematical model it is demonstrated that what appears to be a cohort effect may be a period effect, and conversely. It is recommended that suicide rates be examined for both period and cohort effects before conclusions are drawn regarding trends, and that the complete data set be used for this purpose.