Using Bayesian graphical models to model biases in observational studies and to combine multiple sources of data: application to low birth weight and water disinfection by-products

Summary.  Data in the social, behavioural and health sciences frequently come from observational studies instead of controlled experiments. In addition to random errors, observational data typically contain additional sources of uncertainty such as missing values, unmeasured confounders and selection biases. Also, the research question is often different from that which a particular source of data was designed to answer, and so not all relevant variables are measured. As a result, multiple sources of data are often necessary to identify the biases and to inform about different aspects of the research question. Bayesian graphical models provide a coherent way to connect a series of local submodels, based on different data sets, into a global unified analysis. We present a unified modelling framework that will account for multiple biases simultaneously and give more accurate parameter estimates than standard approaches. We illustrate our approach by analysing data from a study of water disinfection by-products and adverse birth outcomes in the UK.     

Digitalisierung und Konvergenz von Online- und Offline-Welt

The increasing diffusion of mobile phones with internet access (Smartphones) enables more and more consumers to use the mobile internet. In addition, there is a continuing integration of location-based services (LBS). By means of Global Positioning Systems or WiFi-triangulation LBS provide context-aware information to consumers. This leads to a convergence of online and offline worlds. The usage of LBS delivers additional information to consumers (e.g. alternative offers or detailed product information). Therefore LBS do have an influence on consumer behavior. Particularly during the search process, information about prices or geographic distances, that are relevant for the purchase, are of importance. This study analyzes the relevance of location-based internet search empirically. Search costs are estimated on an individual level in a choice-based conjoint analysis using two different products. As a result, location-based internet search is considered to be very relevant for search and influential on consumer behavior. The study shows different consumers preferences and different search costs depending on the product. To conclude the study, the implications are discussed. The major contribution of this study is that it shows that offline and online search do have a mutual impact on each other. Furthermore, search costs are measured in a mobile context.     

The interplay between psychometric and sociometric data and the willingness to adopt innovations

Due to high failure rates of new products, marketers are seeking ways to actively manage the diffusion of innovations. It is well accepted that individuals, who exert a remarkably strong influence on others and are usually called opinion leaders, play an important role in the diffusion process. In marketing, such opinion leaders are typically identified by self-designation, which can be biased. A second option to identify opinion leaders is the use of sociometric data which becomes increasingly available. In this paper we examine the effect of social position on the perceived opinion leadership and the resulting behavioral intention in terms of adoption behavior. In an empirical study, we examine the adoption of a new drug called “Byetta” and find that sociometric opinion leadership is an antecedent of the psychometric characteristic. Our results also confirm that such opinion leaders are more likely to adopt innovations. We also find that opinion seekers, individuals who often sought information or opinions from interpersonal sources, can be more easily identified with psychometric data. They are more restrictive in terms of adoption. Interestingly, both types are active as market mavens and thus may either advocate the adoption of the innovation or may be frankly conservative regarding the new product adoption. If opinion seekers can, however, be persuaded to adopt the innovation, they may also serve as an important multiplier. This has several implications for pharmaceutical firms that try to optimize their sales force—influencing opinion leaders may indeed start and fasten the adoption process of innovative products. But it is also necessary to allay opinion seekers’ doubts since they might slow down the adoption process. The identification of opinion leaders might not only be based on difficult to obtain psychometric data, but also on sociometric data from social networks, which is becoming more and more available.     

Birth order and health of newborns

We examine birth order differences in health of newborns and follow the children throughout childhood using high-quality administrative data on individuals born in Denmark between 1981 and 2010. Family fixed effects models show a positive and robust effect of birth order on health at birth; firstborn children are less healthy at birth. During earlier pregnancies, women are more likely to smoke, receive more prenatal care, and are more likely to suffer a medical pregnancy complication, suggesting worse maternal health. We further show that the health disadvantage of firstborns persists in the first years of life, disappears by age seven, and becomes a health advantage in adolescence. In contrast, later-born children are throughout childhood more likely to suffer an injury. The results on health in adolescence are consistent with previous evidence of a firstborn advantage in education and with the hypothesis that postnatal investments differ between first- and later-born children.