I am fitting a model of a continuous response against 7 fixed effects. When running the Standard Least Squares; my residuals plots/assumptions checks don't look too bad but the Durbin Watson test is suggesting an AR(1) structure in the residuals. This is expected for this particular analysis and is consistent with other data sets looking at the same response/factors. The Time Series platform also suggests an AR(1) dependency in both the Response and Residual columns. However, when I run a Mixed Model in JMP 15 Pro with an AR(1) covariate structure, the results are identical to the Standard Least Squares and the covariance parameter estimates for the AR(1) are all zeros. Unfortunately this analysis is on a classified network so I cannot show model report outputs. All my other analyses on other data sets assign an estimate/std error to the AR(1) covariance parameter estimates and definitely change the values of my t-tests/F-ratios of my fixed effects. How do I explain that all the indicators show an AR(1) covariate structure but the results are identical to the Standard Least Squares estimates? Thank you!
