Why do we need to employ Bayesian statistics and how can we employ it in studies of moral education?: With practical guidelines to use JASP for educators and researchers

Abstract

In this paper, we discuss the benefits of and how to utilize Bayesian statistics in studies of moral education. To demonstrate concrete examples of the applications of Bayesian statistics to studies of moral education, we reanalyzed two datasets previously collected: one small dataset collected from a moral educational intervention experiment, and one big dataset from a largescale Defining Issues Test-2 survey. Results suggest that Bayesian analysis of datasets collected from moral educational studies can provide additional useful statistical information, particularly that associated with the strength of evidence supporting alternative hypotheses, which has not been provided by the classical frequentist approach focusing on P-values. Finally, we introduce several practical guidelines pertaining to how to utilize Bayesian statistics, including the utilization of newly developed free statistical software, Jeffrey's Amazing Statistics Program (JASP), and thresholding based on Bayes Factors, to scholars in the field of moral education.

Evaluating Alternative Correction Methods for Multiple Comparison in Functional Neuroimaging Research

“…In addition to the conventional examination of p -values, we examined Bayes Factors (BF) to compare models and test each effect. According to the suggested statistical guidelines, we considered 2 log BF < 2 as evidence “not worth more than a bare mention,” 2 ≤ 2 log BF < 6 as “positive” evidence, 6 ≤ 2 log BF < 10 as “strong” evidence, and 2 log BF > 10 as “very strong” evidence supporting our hypothesis [42,43]. Bayesian inference was utilized to examine directly the strength of evidence supporting our hypothesis instead of using p -values and to find the best model predicting outcomes among all possible candidate models.…”

“…Bayesian inference was utilized to examine directly the strength of evidence supporting our hypothesis instead of using p -values and to find the best model predicting outcomes among all possible candidate models. In fact, recent debates about statistical testing have warned about using p -values and recommend the utilization of additional testing methods [43,44,45,46]. After conducting the ANOVAs, we conducted classical and Bayesian post-hoc tests to examine inter-condition differences.…”

Implementation of Bayesian Multiple Comparison Correction in the Second-Level Analysis of fMRI Data: With Pilot Analyses of Simulation and Real fMRI Datasets Based on Voxelwise Inference

“…Then, the script numerically estimated x that suffices (10) once α (e.g., α = .05) was provided (see lines Bayesian thresholding was performed by assessing whether a calculated BF value in a specific voxel exceeded a pre-determined BF threshold to determine whether the voxel was significantly activated. Following the guideline that was provided in Kass and Raftery (1995) and was used in previous studies that implemented the similar t-test method Han, Park, & Thoma, 2018;Wagenmakers et al, 2018), we extracted voxels that reported BF greater than 3. According to the guideline and prior research, BF ≥ 3 indicates presence of evidence that positively supports a significant effect (or difference).…”

Applying the Deep Learning Method for Simulating Outcomes of Educational Interventions

“…Recently, there have been debates about whether the conventional threshold for p values, p < .05, is a reliable and valid threshold for statistical decision-making [31]. In particular, we acknowledge that Bayesian inference provides a more reliable way of testing the strength of evidence supporting hypotheses in educational research, compared to the null hypothesis significance testing [15]. Hence, in addition to p values indicating significance in conventional correlation analysis, we refer to Bayes factor (BF) values from Bayesian correlation analysis, indicating the strength of evidence supporting an alternative hypothesis (H 1 : the presence of an actual difference) and the null hypothesis (H 0 : the absence of a difference).…”

“…These issues with traditional statistical modeling have led to a "replication crisis," especially in psychology and medicine, due to lack of replicability of findings [14]. The solutions proposed to address this crisis include more stringent protocols and conventions for research [13], for example, Bayesian statistics and meta-analysis [15,16], and use of predictive machine learning methods [11].…”