This investigation was carried out on annual amounts of wheat production from 65 provinces in seven geographical regions of Turkey during the years 1982 and 1999 to test performance of candidate covariance structures such as: Compound Symmetry (CS), Heterogenous Compound Symmetry (CSH), Unstructured (UN), Huynth Feldth (HF), First-Order Autoregressive (AR(1)), Heterogenous First-Order Autoregressive (ARH(1)), First-Order Ante-Dependence ANTE(1), Toeplitz (TOEP) and Heterogenous Toeplitz (TOEPH) specified for the missing repeated measures data with two fixed factors, viz., region (7 levels) and year (18 levels) using mixed model. In the generation of three missing data sets, deletion operations for the wheat production data were performed at three proportions (10%, 20%, and 30%) on the basis of Missing Completely at Random (MCAR), regardless of any factors under the assessment. The covariance structures were compared with Akaike’s Information Criterion (AIC), Shwartz’s Bayesian Criterion (SBC), and Corrected Akaike’s Information Criterion (AICC) criteria. It was determined that CS was the covariance that produced the best fit among candidate covariance structures with -1158, -1153, -1158 at 10(%) missing data; -916, -912, and -916 at 20(%) missing data and -696, -692, and -696 at 30(%) missing data for AIC, SBC, and AICC, respectively. In conclusion, the investigation results illustrated that several covariance structures in the MIXED procedure of SAS program were easily specified for the repeated measures analysis of the missing data with more time levels.