Een schattingsproef

An experiment on estimation. On ten pieces of card board a line was drawn 100 mm in length and divided by a vertical dash (the pointer) in two parts, in each case at a different point. 40 persons were asked to estimate the length of the left hand portion after they had been told that the total length was 100 mm. The problem discussed is how to decide which person is the ‘best’ estimator. Three criteria are proposed: (I) the sum of the absolute values of the deviations; (II) a similar weighted sum, each deviation being multiplied with the inverse of its standard deviation, in order to account for the obvious differences in accuracy for different positions of the pointer; and (III) by adding for each of the ten readings of one person the proportions of the observers whose readings were worse. Applied to a set of four observers all three criteria indicate No. 20 as the ‘best’ and No. 5 as the ‘worst’. If however we judge the observers by their mutual correlation coefficients, according to a method of Spearman, No. 20 turns out to be the worst. Hence in applying Spearman's criterion caution seems required. It is pointed out that there does not as yet exist a universally recognized and accepted method for determining the relative quality of a set of observers.