High frequency body mass measurement, feedback, and health behaviors

Estimating Multilevel Models on Data Streams

Social scientists are often faced with data that have a nested structure: pupils are nested within schools, employees are nested within companies, or repeated measurements are nested within individuals. Nested data are typically analyzed using multilevel models. However, when data sets are extremely large or when new data continuously augment the data set, estimating multilevel models can be challenging: the current algorithms used to fit multilevel models repeatedly revisit all data points and end up consuming much time and computer memory. This is especially troublesome when predictions are needed in real time and observations keep streaming in. We address this problem by introducing the Streaming Expectation Maximization Approximation (SEMA) algorithm for fitting multilevel models online (or "row-by-row"). In an extensive simulation study, we demonstrate the performance of SEMA compared to traditional methods of fitting multilevel models. Next, SEMA is used to analyze an empirical data stream. The accuracy of SEMA is competitive to current state-of-the-art methods while being orders of magnitude faster.

Childhood Adversity and Adult Health: The Role of Developmental Timing and Associations With Accelerated Aging

Childhood adversity has been associated with poor adult health. However, it is unclear whether timing of adversity matters in this association and whether adversity is related to poorer age-related physical health status. A representative sample of the adult Dutch population ( N = 3,586, age M = 54.94, age range = 18-92) completed surveys on health and diagnoses of age-related diseases. Information about weight and fat percentage was collected using weighing scales and childhood experiences were assessed retrospectively. Adversity was associated with higher body mass index and fat percentage, more physical problems, and high cholesterol, and this association was most pronounced in individuals with experiences of adversity during early adolescence. In addition, individuals with adversity more often reported physical problems or a medical diagnosis at a younger age. This study indicates that (1) timing of exposure to adversity matters in the relationship between experienced childhood adversity and health and (2) adversity is associated with a higher prevalence of age-related diseases at earlier ages.

To what extent are members of an online panel willing to share different data types? A conjoint experiment

Recently, the idea of ‘data fusion’, that is, of combining different types of data, became quite popular because of the advances of new technologies. In particular, several studies started investigating the possibility of combining survey data with other data types in order to get a more complete or accurate picture of the reality and/or to reduce survey burden. One key element, then, is the willingness of people to share different types of data, beyond survey answers. In this article, we investigate to what extent members from an opt-in online panel in Spain are willing to share different types of information that have in general not been studied before in the literature: records of their surrounding sound (audiotracking), information from their email inbox (in different ways, sharing the email credentials, using an email plug-in or redirecting emails, partially or totally), sensorial reactions measured by a wearable device (neuroscience) and public information about them available online. We use a choice-based conjoint analysis in order to study the level of willingness depending on the incentives offered in exchange, and we present the level of willingness by gender and age groups. Overall, we find huge differences in the level of willingness across data types. Increasing the incentives, on the contrary, does not improve the willingness so much, even if there is a positive trend. Some differences are observed across gender and age groups but most of them are not statistically significant.

You can be too thin (but not too tall): Social desirability bias in self-reports of weight and height

Previous studies of survey data from the U.S. and other countries find that women tend to understate their body weight on average, while both men and women overstate their height on average. Social norms have been posited as one potential explanation for misreporting of weight and height, but lack of awareness of body weight has been suggested as an alternative explanation, and the evidence presented to date is inconclusive. This paper is the first to offer a theoretical model of self-reporting behavior for weight and height, in which individuals face a tradeoff between reporting an accurate weight (or height) and reporting a socially desirable weight (or height). The model generates testable implications that help us to determine whether self-reporting errors arise because of social desirability bias or instead reflect lack of awareness of body weight and/or other factors. Using data from the National Health and Nutrition Examination Survey (NHANES) from 1999 to 2010, we find that self-reports of weight offer robust evidence of social desirability bias. However, lack of awareness of weight may also contribute to self-reporting biases, and this factor appears to be more important within some demographic groups than others. Among both women and men, self-reports of height exhibit significant social desirability bias only among those of below-average height, and very few individuals underreport their height. Implied self-reports of BMI exhibit gender-specific patterns similar to those observed for self-reporting of weight, and the inferred social norms for BMI (20.8 for women and 24.8 for men) are within the "normal" range established by public health institutions. Determining why individuals misreport their weight has important implications for survey design as well as for clinical practice. For example, our findings suggest that health care providers might take additional steps to increase self-awareness of body weight. The framework also helps to explain previous findings that the degree of self-reporting bias in weight is stronger in telephone surveys than it is in in-person surveys.