[EMAIL PROTECTED] (Paul R Swank) wrote in message news:<[EMAIL PROTECTED]>... > And doing a Pearson Coorelation and a t-test doesn't tell you the overall > impact of the error.
Let X and Y denote the two tests -- test-retest, alternate forms/raters, whatever. Then a measure of the "total impact" of unreliability is the mean square difference, E(X-Y)^2, which may be decomposed as follows: E(X-Y)^2 = (Mx-My)^2 + (Sx-Sy)^2 + 2*Sx*Sy(1-Rxy), in which the three components are easily calculated and compared, and are due to a difference in level, a difference in spread, and lack of correlation. . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
