Differences between complete and incomplete couples in physical health research: Implications for methods and generalizability.