Modeling time-series count data: the unique challenges facing political communication studies

Area of Science:

• Political Science
• Communication Studies
• Quantitative Methods

Background:

• Scholars in media and politics often analyze time-series count data, such as issue coverage over time.
• While time dependence and dynamic causality are key considerations, the count nature of the outcome variable is frequently overlooked.
• This oversight is particularly problematic when analyzing data with low counts.

Purpose of the Study:

• To highlight the importance of appropriate model specification for time-series count data in political communication research.
• To advocate for the Poisson autoregressive model as a suitable method for media studies.
• To demonstrate how incorrect model assumptions can lead to invalid research inferences.

Main Methods:

• Replication of existing studies (Flemming et al., 1997; Peake and Eshbaugh-Soha, 2008; Ura, 2009) using the Poisson autoregressive model.
• Comparison of model inferences when assumptions are met versus when they are violated.
• Utilizing a simulation procedure to dynamically illustrate model effects and uncertainty estimates.

Main Results:

• Models that fail to account for the count nature of the data and its specific assumptions can produce invalid inferences.
• The Poisson autoregressive model is demonstrated to be a robust approach for analyzing time-series count data in this field.
• Dynamic illustrations with uncertainty estimates enhance the understanding of model-based findings.

Conclusions:

• Correct model specification, specifically employing the Poisson autoregressive model, is vital for valid analysis of time-series count data in political communication.
• Researchers must consider the distributional properties of count data to avoid erroneous conclusions.
• The findings have direct implications for the practical application of statistical models in media and politics research.