|
Abstract:
|
To study the evolution of media narratives that emerge in the aftermath of mass shooting events, this proposed project aims to develop a procedure for synthesizing evidence from separate qualitative content analyses of media articles following mass shooting events using modern Bayesian methodologies such as multivariate multinomial probit (MVMNP) regression for categorical data and calibration-based multiple imputation. The availability of machine learning-based text analysis programs that can automate the coding process for large numbers of media articles coupled with human coders that can review a subsample of those articles in more detail gives rise to viewing the coding problem as a missing data problem. Applying multiple imputation to the incomplete dataset would yield plausible codes for the unreviewed articles while properly accounting for the prediction uncertainty, providing a statistically secure foundation for the comparison of media narratives across events.
|
