Network Autocorrelation Models
Model Parameter Estimation and Statistical Properties
We seek to understanding the fundamental estimation problems of this model. For example, estimation bias of rho or multiple rho’s under different circumstances, the statistical power to detect in detecting rhos and betas as a function of various network characteristics.
The Multi-Behavior Multi-Network (MBMN) Model
Naturalistically and realistically, human beings are living in a well-connected social world and engaging in a complex array of multiple behaviors daily. However, scientifically and academically, social scientists have been typically trained to aim at understanding and predicting a single behavior of an individual in relatively isolated settings. For example, in the health domain, researchers have been working hard to find magic-bullet type of resolutions for specific health problems such as smoking, drug using, any other risky behaviors, etc. In organizational psychology, scholars often focus on one behavior of their research interest –for example, task performance, lateness, work withdrawal, turnover, etc – and try to understand the behavior separately through either individual characteristics such as intelligence and personality or situational characteristics such as leadership and organizational climate. As such, these conventional approaches have unfortunately ignored two critical natures of human behaviors: (1) clustering, and (2) susceptibility to social influence through networks through which one is socially connected.
Indeed, human achieves goals through a cluster of behaviors instead of single one behavior. For example, a healthy life style not only requires a reduction in drug or alcohol or tobacco
use, but also requires behaviors of excising, eating well, and reducing psychological stress. A successful employee at workplace not only engages in task performance, but also organizational citizenship behaviors, as well as disengages in any counterproductive work behaviors. Thus, understanding the cluster of related behaviors is necessitated for advanced behavioral research in the decade to come. In addition, these clusters of behaviors are under tremendous influence of various social networks through network diffusion and contagion, which defined as one’s connected nodes (e.g., friends, coworkers) influence one’s own opinions, emotions, and behaviors. To accommodate these two critical behavior natures, our Lab proposes a novel method: Multi-Behavior Multi-Network Modeling, that takes the advantage of big date computation and aims to develop a complex computing model for advanced research in many social sciences, and then preliminarily applies this advanced modeling method into health and organizational domains to advance our current understanding of behaviors in these two domains.