Testing and Estimation of Social Network Dependence With Time to Event Data

Area of Science:

• Social Network Analysis
• Statistical Modeling
• Computational Social Science

Background:

• Events spread rapidly across social networks, prompting interest in how social connections influence individual behavior.
• Understanding social network dependence is crucial for predicting responses to events, such as game adoption.

Purpose of the Study:

• To propose a novel latent spatial autocorrelation Cox model for analyzing social network dependence with time-to-event data.
• To introduce a latent indicator to assess the impact of friends' features on an individual's survival time.

Main Methods:

• Development of a score-type test to detect social network dependence.
• Implementation of an EM-type algorithm for parameter estimation when dependence exists.
• Application of the model to a time-to-event dataset on mobile game engagement.

Main Results:

• The proposed model effectively studies social network dependence in time-to-event data.
• Simulation studies and a real-world case study demonstrate the performance of the developed test and estimators.
• The findings provide insights into how social influence affects user engagement with online platforms.

Conclusions:

• The latent spatial autocorrelation Cox model offers a robust framework for quantifying social network dependence.
• The developed statistical methods are valuable for analyzing complex social behaviors in online environments.
• This research contributes to the emerging field of social network dependence analysis.