THE ROLE OF SUB-CLASSIFICATION IN STATISTICAL ANALYSIS.

The article presents comments of the author on the article "Is Prediction Feasible in Social Work? An Inquiry Based Upon a Sociological Study of Parole Records," by Ernest W. Burgess. Burgess' paper seems to me a valuable contribution. It may be regarded as a discussion of the central factor of scientific administration; that is to say, of supervision or control when that supervision ceases to be arbitrary, empirical and rule-of-thumb. While any one who does any work of any kind would naturally prefer to be the sole judge of the quality of that work, it is coming to be recognized that it is necessary to have any job appraised by other people than those who do it. If this work of appraisal by professional colleagues or by others in positions of responsibility is to be done on a rational basis, it is necessary to find standards of performance based on an analysis of past experience under the same or similar conditions. The question can then be asked, are the results actually secured by a particular worker better or worse than those previously secured, or secured elsewhere under similar conditions? Further, are the results of method A better than those of method B? In the field of business, great progress has beta made in recent years in determining and using such standards of performance where the output is tangible, made up of easily measured parts.