<p>Big data is ubiquitous in practice and often imposes heavy computational burdens. To reduce computational costs and ensure efficient parameter estimation, an optimal subsampling method is proposed for estimating parameters in marginal models with massive longitudinal data. The proposed method does not require assumptions about the joint distribution or the correlation structure of the response variables. The optimal subsampling probabilities are derived, and the consistency and asymptotic normality of the proposed estimator are established. Extensive simulation studies are carried out to evaluate the performance of the proposed method for continuous, binary, and count data, using four different working correlation matrices. The simulation results demonstrate that the proposed method significantly outperforms the uniform subsampling approach. Furthermore, a depression dataset is analyzed to illustrate the practical application of the proposed method.</p>