Increments-decrements Analysis - Impact of Assumption Violation on Empirical Rate of Type-i Errors

Increments-decrements technique was investigated in an extensive Monte Carlo simulation study. It was demonstrated that violated assumptions yield problems of inference in the form of important differences between exact and empirical type I error rates. Skewness in random error of first order autoregressive model affects dramatically empirical type I error rate. It was found that the empirical type I error was affected by nonzero serial dependence, although this technique showed an acceptable robustness for levels of autocorrelation near zero. Extreme caution is recommended for use this strategy in A-B designs where autocorrelated series are not unusual.