An experimental factor analysis of cancer mortality in England and Wales 1921–30

Abstract

A Factor Analysis has been made of the co-variation between the mortality rates from cancer of ten male body sites and of eight female body sites, in thirty large towns in England and Wales from 1921 to 1930. The method of analysis adopted is Hotelling's method of Principal Components.Four male and four female Factors are obtained, which together account for approximately three-quarters of the total variance.A Factor is found to be associated with cancer of the larynx, oesophagus, stomach and tongue in men and with cancer of the stomach and negatively with cancer of the breast and ovaries in women. In both sexes, the Factors are associated with an index of adverse social conditions.Another Factor is found to be associated with cancer of the rectum and bladder in men and this Factor is associated with good social conditions.A special Factor associated with cancer of the colon is unrelated to the mortality from cancer of other sites, save that in women it is negatively associated with cancer of the rectum.A Factor for cancer of the lung in men is unrelated to cancer of the larynx, and is inversely related to cancer of the tongue.We are most grateful to Dr Percy Stocks, late of the General Register Office, and to Dr W. P. D. Logan, Chief Medical Statistician of the General Register Office, for their help in providing us with the relevant basic material; and to Miss E. M. Hines, Miss A. H. Huntley and Miss M. Rogers for assistance in the calculations.