Independent component analysis (ICA), a blind source separation method, has been shown to be a useful approach to identify genetic components representing combined effects from multiple mutations. However, the ICA order selection for genotype data has been a challenge, since a genetic component usually accounts for a small amount of variance of the data, and makes it difficult to distinguish true signals from background. To address this issue, we propose to select ICA order based on consistency and implement three strategies in this study. Simulations demonstrate robust performances of all three strategies where the selected orders lead to optimal results regardless of ICA performances.