This study aims to evaluate and compare the accuracy of two imputation methods, VAR-IMMA (Vector Autoregressive - Moving Average Imputation Method) and MICE PMM (Multiple Imputation by Chained Equations with Predictive Mean Matching), for handling missing data in multivariate time series. The evaluation was conducted using synthetic time series data generated from a stable VAR(1) model. Missing values were introduced under the Missing at Random (MAR) mechanism across five different missingness levels: 5%, 10%, 15%, 20%, and 25%. Each scenario was repeated 100 times. After imputation, each completed dataset was fitted using a VAR model, and the fitting accuracy was assessed by comparing the fitted values with the original data using RMSE and MAPE as performance metrics. The results showed that VAR-IMMA consistently outperformed MICE PMM across all scenarios, particularly at lower levels of missingness. These findings suggest that VAR-IMMA is more effective in preserving the temporal and inter-variable relationships in multivariate time series with MAR missing data.