A Time-Varying Dynamic Partial Credit Model to Analyze Polytomous and Multivariate Time Series Data

Area of Science:

• Psychological research
• Statistical modeling
• Time series analysis

Background:

• Electronic devices and statistical methods enhance individual-level psychological process research.
• Existing models struggle with complex data, such as Likert-scale items and nonstationary time series.
• Ignoring variable scale and stationarity assumptions can bias results.

Purpose of the Study:

• To propose a novel statistical model for analyzing complex psychological data.
• To address limitations of existing methods in handling polytomous data and nonstationary time series.
• To introduce the time-varying dynamic partial credit model (TV-DPCM).

Main Methods:

• Combined the partial credit model (PCM) from item response theory with the time-varying autoregressive (TV-AR) model.
• Developed the time-varying dynamic partial credit model (TV-DPCM).
• Tested the TV-DPCM's performance and accuracy via a simulation study.

Main Results:

• The TV-DPCM appropriately analyzes multivariate polytomous data.
• The model effectively handles nonstationary time series in psychological dynamics.
• Simulation study confirmed the model's performance and accuracy.

Conclusions:

• The TV-DPCM offers a robust solution for analyzing complex psychological time series data.
• The model accommodates Likert-scale measurements and dynamic, nonstationary processes.
• Demonstrated practical application and interpretation of the TV-DPCM with empirical data.