In every case the correlation between grades and salary is positive, although the coefficients are all small. This means that in the long run there is a general tendency for the good salaries to go to the men whose grades were high, but that there are many exceptions to the rule. Certainly in no class is the opposite tendency shown, for the good salaries to go to the poor students. It is probable that the correlations found here are as low as they are partly because in this technical school there is no special effort made to encourage high grades for their own sake, the emphasis being rather on getting a good average rating. TABLE 18 SHOWING THE CORRELATION BETWEEN SCHOOL STANDING AND SALARIES EARNED IN LATER LIFE (RICE) See Text for Explanation ------------------------------------------------------------ | | Correlation by | Correlation by Class and Year| Cases | Pearson Method,| Per Cent of Unlike | | and P.E. | Signs, and P.E. ---------------+-------+----------------+------------------- Mechanical '07 | 35 | .36 .08 | .22 .09 Mechanical '08 | 41 | .25 .09 | .34 .08 Mechanical '09 | 39 | .21 .09 | .06 .10 | | | Electrical '07 | 26 | .16 .13 | .25 .12 Electrical '08 | 36 | .46 .08 | .51 .08 Electrical '09 | 41 | .16 .10 | .28 .09 | | | Averages | | .267 | .277 ------------------------------------------------------------ Just what these degrees of correlation mean is made somewhat more apparent if we treat the data in another way. If instead of computing coefficients of correlation we divide each class of men into four quartiles, and determine the average salaries of the men in these quartiles, we get very definite results. The upper quartile or group will now contain that fourth of the class whose grades were highest. The second, third and fourth quartiles will in turn represent decreasing degrees of academic proficiency. If the average salaries are the same for all quartiles, this will mean that there is no relation between salary and school grades. But if the salary varies with the grades, this will be a significant result. The actual data are as follows: TABLE 19 RICE'S DATA PRESENTED IN A REVISED FORM