Generative adversarial networks (GANs) are implicit generative models that can be used for data imputation as an unsupervised learning problem. This works introduces an iterative GAN architecture for data imputation based on the invertibility of the generative imputer. This property is a sufficient condition for the convergence of the proposed GAN architecture. The performance of imputation is demonstrated by applying different imputation algorithms on the traffic speed data from Guangzhou (China). It is shown that our proposed algorithm produces more accurate results compared to those of previous GAN-based imputation architectures.