Abstract
To examine nonresponse effects on variable relationships, data from a panel study of 1,870 older Iowa residents are reported. When data are accumulated by wave of response, differences are minor on variable distributions (percentage, mean, standard deviation), bivariate relationships (correlation), and multivariate relationships (regression), suggesting that the assumed impact of nonresponse bias on variable relationships has been overstated.