An empirical examination of the relationship between nonresponse rate and nonresponse bias