Associations with social networks, social support, health status and psychiatric morbidity in three samples of elderly people
Using longitudinal data to estimate nonresponse bias