Mental Computations of Ideological Choice and Conviction: The Utility of Integrating Psycho-Economics and Bayesian Models of Belief

A quick scan of the political landscape reveals that people differ in the ideologies they embrace and advocate. Why do individuals prefer certain ideologies over others? A formal analysis of psychological needs and consumption desires suggests that it is possible to compute the subjective utility of selecting one ideology over another, as though it were a purchasing decision. Given resources, constraints, and available options, individuals can rationally choose the ideology that best matches or resonates with their interests. It is a compelling framework that can take into account how diverse ideologies satisfy people’s diverse and multidimensional psychological and material needs. This psycho-economic model is ambitious and informative, and I will argue that it can be even more encompassing and enlightening if it is expanded to incorporate two critical components of ideological cognition: (1) the nature of ideological conviction and extremism and (2) the dynamic, probabilistic mental computations that underlie belief formation, preservation, and change. Firstly, I will argue that a formal model of ideological choice cannot escape the question of the strength of ideological commitment. In other words, we need to ask not only about which ideologies individuals choose but also about how strongly they adhere to these ideologies once those are chosen. An analysis of ideological choice needs to be accompanied by an analysis of ideological conviction. Secondly, in order to build a robust sense of the rationality behind ideological thinking, it is useful to incorporate principles of uncertainty and probability-based belief updating into the formal model of ideological worldviews. Bayesian models highlight how human brains seek to build predictive models of the world by updating their beliefs and preferences in ways that are proportional to their prior expectations and sensory experiences. Consequently, incorporating Bayesian principles into the formal model of ideological choice will provide a more wholistic understanding of what happens when a mind enters the market for belief systems – and why a mind can, at times, purchase toxic doses of the ideologies that sellers and entrepreneurs offer on display.

Bayesian paired comparison with the bpcs package

This article introduces the R package (Bayesian Paired Comparison in Stan) and the statistical models implemented in the package. This package aims to facilitate the use of Bayesian models for paired comparison data in behavioral research. Bayesian analysis of paired comparison data allows parameter estimation even in conditions where the maximum likelihood does not exist, allows easy extension of paired comparison models, provides straightforward interpretation of the results with credible intervals, has better control of type I error, has more robust evidence towards the null hypothesis, allows propagation of uncertainties, includes prior information, and performs well when handling models with many parameters and latent variables. The package provides a consistent interface for R users and several functions to evaluate the posterior distribution of all parameters to estimate the posterior distribution of any contest between items and to obtain the posterior distribution of the ranks. Three reanalyses of recent studies that used the frequentist Bradley–Terry model are presented. These reanalyses are conducted with the Bayesian models of the package, and all the code used to fit the models, generate the figures, and the tables are available in the online appendix.

Bayesian Linear Regression: Teaching resources and software

Bayesian models are very important in modern data science. These models can be used to derive estimatesfor noisy and sparse data. This manuscript outlines the basics and derivations of a Bayesian linearregression model. Source code for performing Bayesian linear regression is also provided. I hope thisresource will enable broader understanding of the basics of Bayesian models.

63. PK-RNN-V: A Deep Learning Model for Vancomycin Therapeutic Drug Monitoring using Electronic Health Record Data

Background Therapeutic drug monitoring (TDM) for vancomycin (VAN) with Bayesian models is recommended by national guidelines. However, limited data incorporating the models may hurt the performance. Our aim is to develop a novel deep learning-based pharmacokinetic model for vancomycin (PK-RNN-V) using electronic medical records (EHRs) data to achieve more accurate and personalized predictions for VAN levels. Methods EHR data were retrospectively retrieved from Memorial Hermann Hospital System, comprising 14 hospitals in the greater Houston area. All patients who received VAN and had any VAN levels were eligible. Patients receiving hemodialysis and extracorporeal membrane oxygenation were excluded. Demographic data, vital signs, diagnostic codes, concomitant medications, VAN administration, and laboratory data were preprocessed as longitudinal data. VAN infusion, VAN level measurement, or each hospital day were the time steps for the models. The dataset was splited 70:15:15 for training, validation, and test sets. Our PK-RNN-V model predicted individual patient volume distribution (v) and VAN elimination (k) at each time step using an irregular timesteps GRU model. To compare, Bayesian models were developed from publicly available models, and tuned to feedback the first VAN level to update the v and k. (VTDM) Results A total of 12,258 patients with 195,140 encounters were identified from Aug, 2019 and March, 2020. After exclusion of 6,775 patients, 5,483 patients with 8,689 encounters were included. Table 1 summarized the characteristics of patients included in our study. 55,336 doses of VAN were administered with a median dosage of 1.0 gm. VAN levels were measured 18,588 times at various timings. The median VAN level was 14.7 mcg/mL Table 2 described the performance of our models and VTDM models. Our model exhibited better performance compared to VTDM model (RMSE: 5.64 vs. 6.57, respectively). Figure 1 shows example prediction curves of VAN levels from each model. Conclusion PK-RNN-V model is a novel approach to predict patient PK and VAN levels. Our results revealed promising performance of this model. Our model can take a wide range of real-world patient data into the model. Further studies are warranted for external validations and model optimizations. Disclosures All Authors: No reported disclosures