In recent years, brain network-oriented analyses have become increasingly important in neuroimaging studies to study neural circuitry among healthy brains and also to investigate network differences between subpopulations. These analyses often encounter challenges including low signal-to-noise ratio in neuroimaging data, high between-subject heterogeneity in brain networks and variations in images and networks on the same subject across different sessions. In this talk, we present several new statistical methods to help improve the reliability and reproducibility in brain network analysis. The proposed methods tackle the aforementioned challenges by joint network learning using brain images obtained on the same subjects under different conditions, by multimodality integrative network modeling that incorporates anatomical structure in constructing brain functional networks, and by developing novel statistical tests to more accurately detect network differences between subpopulations. We will discuss the theoretical properties of the methods and demonstrate their performance through simulation studies and real-world neuroimaging data examples.