A survey increases responding households' weights to represent similar households that did not respond. What has this procedure done to the missing households' individual questionnaire answers?