Associations between social support and depressive symptoms: social causation or social selection—or both?