Some Effects of Errors of Measurement on Multiple Correlation

摘要: Abstract In many multiple correlation and regression studies the values of variables are difficult to measure accurately. This paper attempts discuss effects such errors measurements on squared coefficient. With independent a multivariate normal model, effect is very roughly that R 2 becomes reduced about , where gy coefficient reliability y weighted mean coefficients x's. If most intercorrelations among true X's detrimental 2, value R'2 may exceed by 10 percent 20 percent. The situation in which correlated with interpretation briefly discussed.