MEASUREMENT OF CHANGE AND THE LAW OF INITIAL VALUES - A COMPUTER-SIMULATION STUDY

Computer simulations were used to examine the effect of a negative correlation between baseline and change on two measures of change: gain scores and covariance. When a negative correlation resulted from decreasing variance from pretest to posttest, as might be due to the presence of floors or ceilings (the Law of Initial Values) or skewness, gain scores were less powerful than convariance. However, when a negative correlation resulted from measurement error alone, the relative power of gain scores and covariance was unaffected. It was concluded that the correlation between baseline and change, by itself, does not invalidate the use of gain scores. However, when the negative correlation is accompanied by a decrease in variance from pretest to posttest, covariance is a superior index of change.