Non-response weighting is a commonly used method to adjust for bias due to unit nonresponse in surveys. Theory and simulations show that, to reduce bias effectively without increasing variance, a covariate that is used for non-response weighting adjustment needs to be highly associated with both the response indicator and the survey outcome variable. In practice, these requirements pose a challenge that is often overlooked, because those covariates are often not observed or may not exist. Surveys have recently begun to collect supplementary data, such as interviewer observations and other proxy measures of key survey outcome variables. To the extent that these auxiliary variables are highly correlated with the actual outcomes, these variables are promising candidates for non-response adjustment. In the present study, we examine traditional covariates and new auxiliary variables for the National Survey of Family Growth, the Medical Expenditure Panel Survey, the American National Election Survey, the European Social 389 390 K r e u t e r e t a l . 1 7 3 ( 2 0 1 0 ) Surveys and the University of Michigan Transportation Research Institute survey. We provide empirical estimates of the association between proxy measures and response to the survey request as well as the actual survey outcome variables. We also compare unweighted and weighted estimates under various non-response models. Our results from multiple surveys with multiple recruitment protocols from multiple organizations on multiple topics show the difficulty of finding suitable covariates for non-response adjustment and the need to improve the quality of auxiliary data. i n J o u r n a l o f t h e ro y a l S t a t i S t i c a l S o c i e t y a

No abstract

Summary Interviewer observations made during the process of data collection are currently used to inform responsive design decisions, to expand the set of covariates for nonresponse adjustments, to explain participation in surveys, and to assess nonresponse bias. However, little effort has been made to assess the quality of such interviewer observations. Using data from the Los Angeles Family and Neighbourhood Survey (L.A.FANS), this paper examines measurement error properties of interviewer observations of neighbourhood characteristics. Block level and interviewer covariates are used in multilevel models to explain interviewer variation in the observations of neighbourhood features.

Background: Although cardiovascular disease (CVD) is the leading cause of mortality in Latin American women, limited data exist on CVD perceptions in this population. This study aimed to assess CVD awareness and knowledge of women from Santiago, Chile. Methods: This was a cross-sectional study conducted in women 35 to 70 years old. A multistage probability sampling (stratified by age and socioeconomic level) was used for participant selection. Participants completed a home survey about knowledge of CVD, risk factors, and perceived risk (based on standardized questions from the American Heart Association awareness survey). Results: 723 women participated in the study (mean age: 51 ± 9 years; 17.6% with high education level). Only 9.3% of the respondents mentioned CVD as women's primary health problem, whereas 22.7% and 16.1%, respectively, listed breast cancer and other cancers. When asked to identify the leading cause of women's death, only 14.4% identified CVD compared to 69.1% who recorded cancer. Older women (≥ 55 years) more likely identified CVD as the main cause of death: (OR 2.9: 95% CI = 1.8-4.5) versus younger women (<55 years). CVD family history was also associated with higher awareness of CVD as the leading cause of death (OR 1.7: 95% IC; p = 1.1-2.6). Instead, women with middle education level were less likely to mention CVD as the main women's killer. Conclusions: Chilean women from Santiago have a low awareness of CVD as the leading cause of death and do not recognize CVD as their prominent health problem. Efforts should focus on increasing awareness and knowledge about CVD especially in young women.

Background We compare self-reported prevalence of drug use and indicators of data quality from two different response modes (with and without an independent answer sheet for recording responses) in a survey conducted in 2015 among secondary school students. Methods Stratified cluster-randomized study conducted among students in grades 8 to 12 from public, private and subsidized schools in Chile (N = 2,317 students in 122 classes). Measurements included were: percentage reporting substance use (tobacco, alcohol, marijuana, cocaine, ecstasy); number of inconsistent responses; number of item nonresponses; percentage of extreme reports of drug use; percentage reporting using the nonexistent drug, relevón; and completion times. Results Compared with those who responded directly in the questionnaire booklet, students who used a separate answer sheet took 17.6 more minutes (95% confidence interval [CI]: 14.4–20.8) to complete the survey and had on average 1.5 more inconsistent responses (95%CI: 0.91–2.14). The prevalence and variance of drug use was higher among those who used an answer sheet for all substances except tobacco; the prevalence ratio (PR) of reported substance use for low-prevalence substances during the past year were: cocaine PR=2.5 (95%CI: 1.6–4.1); ecstasy PR=5.0 (95%CI: 2.4–10.5); relevón PR=4.8 (95%CI: 2.5–9.3). Conclusions Using an answer sheet for a self-administered paper-and-pencil survey of drug use among students result in lower quality data and higher reports of drug use. International comparison of adolescent drug use from school-based surveys should be done with caution. The relative ranking of a country could be misleading if different mode of recording answers are used.