marginally significantp = 0.051
Model 1a (HR ~ time + group + autoregression) fit significantly better than Model 0a (χ 2 (2) = 9.28, p = 0.010), while the difference in the goodness of fit between Model 1a and Model 2a (HR ~ time * group + autoregression) was marginally significant (χ 2 (1) = 3.81, p = 0.051).