“…The use of OLS regression for analyzing multiple-baseline data has been advocated (Huitema & McKean, 1998), but the use of OLS regression methods has also raised concerns, because the errors in the model are assumed to be independent. Many have argued that errors closer in time may be more similar to each other than independently selected errors, and, therefore, the errors may be positively autocorrelated instead of independent (Kratochwill et al, 1974;Matyas & Greenwood, 1997). Furthermore, positive autocorrelation impacts the statistical inferences such that there is a greater chance of Type I errors (Greenwood & Matyas, 1990;Toothaker, Banz, Noble, Camp, & Davis, 1983), which in turn implies that 95% confidence intervals would contain the actual effect less than 95% of the time.…”