Bayesian evaluation of informative hypotheses for multiple populations

Bayesian evaluation of diverging theories of episodic and affective memory distortions in dysphoria

People suffering from dysphoria retrieve autobiographical memories distorted in content and affect, which may contribute to the aetiology and maintenance of depression. However, key memory difficulties in dysphoria remain elusive because theories disagree how memories of different valence are altered. Here, we assessed the psychophysiological expression of affect and retrieved episodic detail while participants with dysphoria (but without a diagnosed mental illness) and participants without dysphoria relived positive, negative, and neutral memories. We show that participants with dysphoria retrieve positive memories with diminished episodic detail and negative memories with enhanced detail, compared to participants without dysphoria. This is in line with negativity bias but not overgeneral memory bias theories. According to confirmatory analyses, participants with dysphoria also express diminished positive affect and enhanced negative affect when retrieving happy memories, but exploratory analyses suggest that this increase in negative affect may not be robust. Further confirmatory analyses showed that affective responses to memories are not related to episodic detail and already present during the experience of new emotional events. Our results indicate that affective memory distortions may not emerge from mnemonic processes but from general distortions in positive affect, which challenges assumptions of memory theories and therapeutics. Protocol registration: The Stage 1 protocol for this Registered Report was accepted in principle on the 18rd of March 2021. The protocol, as accepted by the journal, can be found at https://doi.org/10.6084/m9.figshare.14605374.v1 .

Reduced Thickness of the Retina in de novo Parkinson's Disease Shows A Distinct Pattern, Different from Glaucoma

Background

Parkinson's disease (PD) patients experience visual symptoms and retinal degeneration. Studies using optical coherence tomography (OCT) have shown reduced thickness of the retina in PD, also a key characteristic of glaucoma.

Objective

To identify the presence and pattern of retinal changes in de novo, treatment-naive PD patients compared to healthy controls (HC) and early primary open angle glaucoma (POAG) patients.

Methods

Macular OCT data (10×10 mm) were collected from HC, PD, and early POAG patients, at the University Medical Center Groningen. Bayesian informative hypotheses statistical analyses were carried out comparing HC, PD-, and POAG patients, within each retinal cell layer.

Results

In total 100 HC, 121 PD, and 78 POAG patients were included. We showed significant reduced thickness of the inner plexiform layer and retinal pigment epithelium in PD compared to HC. POAG patients presented with a significantly thinner retinal nerve fiber layer, ganglion cell layer, inner plexiform layer, outer plexiform layer, and outer photoreceptor and subretinal virtual space compared to PD. Only the outer segment layer and retinal pigment epithelium were significantly thinner in PD compared to POAG.

Conclusions

De novo PD patients show reduced thickness of the retina compared to HC, especially of the inner plexiform layer, which differs significantly from POAG, showing a more extensive and widespread pattern of reduced thickness across layers. OCT is a useful tool to detect retinal changes in de novo PD, but its specificity versus other neurodegenerative disorders has to be established.

Bayesian sequential designs in studies with multilevel data

In many studies in the social and behavioral sciences, the data have a multilevel structure, with subjects nested within clusters. In the design phase of such a study, the number of clusters to achieve a desired power level has to be calculated. This requires a priori estimates of the effect size and intraclass correlation coefficient. If these estimates are incorrect, the study may be under- or overpowered. This may be overcome by using a group-sequential design, where interim tests are done at various points in time of the study. Based on interim test results, a decision is made to either include additional clusters or to reject the null hypothesis and conclude the study. This contribution introduces Bayesian sequential designs as an alternative to group-sequential designs. This approach compares various hypotheses based on the support in the data for each of them. If neither hypothesis receives a sufficient degree of support, additional clusters are included in the study and the Bayes factor is recalculated. This procedure continues until one of the hypotheses receives sufficient support. This paper explains how the Bayes factor is used as a measure of support for a hypothesis and how a Bayesian sequential design is conducted. A simulation study in the setting of a two-group comparison was conducted to study the effects of the minimum and maximum number of clusters per group and the desired degree of support. It is concluded that Bayesian sequential designs are a flexible alternative to the group sequential design.

Demand-Resource Profiles and Job Satisfaction in the Healthcare Sector: A Person-Centered Examination Using Bayesian Informative Hypothesis Testing

Work characteristics may independently and jointly affect well-being, so that whether job demands deplete or energize employees depends on the resources available in the job. However, contradictory results on their joint effects have emerged so far in the literature. We argue that these inconsistencies can be partially explained by two arguments in the contemporary literature in the field. First, most studies in the job design domain are based on classic variable-centered methodologies which, although informative, are not well suited to investigate complex patterns of interactions among multiple variables. Second, these studies have mainly focused on generic work characteristics (e.g., workload, control, support), and are lacking in occupational specificity. Thus, to overcome these limitations, in the current research we include generic and occupation-specific work characteristics and adopt a person-centered approach to (a) identify different patterns of interactions of job demands and resources in a sample of healthcare employees, and (b) determine the degree to which these patterns are associated with employee well-being. We involved a sample of 1513 Italian healthcare providers and collected data on key job demands (workload, emotional dissonance, patient demands and physical demands) and resources (control, management support and peers' support). We focused on job satisfaction as a broad indicator of well-being. Latent profile analysis revealed four profiles of job demands and resources: high strain-isolated, resourceless, resourceful and active job on the ward. The results of Bayesian informative hypothesis testing showed the highest support for the hypothesis stating that healthcare employees belonging to the active job on the ward profile (medium-high demands, high resources) were the most satisfied. Conversely, employees belonging to the high strain-isolated profile (high demands, low resources) and the resourceless profile (medium-low demands, low resources) were the least satisfied. Overall, our study confirms the key role played by job resources in determining well-being in high-risk sectors, demonstrating that job satisfaction can develop both in challenging and less demanding situations. On a practical level, mapping the complexity of the healthcare psychosocial work environment has important implications, allowing for a better assessment process of employee well-being and helping to identify the most effective and fitting interventions.

Sample size determination for Bayesian ANOVAs with informative hypotheses

Researchers can express their expectations with respect to the group means in an ANOVA model through equality and order constrained hypotheses. This paper introduces the R package SSDbain, which can be used to calculate the sample size required to evaluate (informative) hypotheses using the Approximate Adjusted Fractional Bayes Factor (AAFBF) for one-way ANOVA models as implemented in the R package bain. The sample size is determined such that the probability that the Bayes factor is larger than a threshold value is at least η when either of the hypotheses under consideration is true. The Bayesian ANOVA, Bayesian Welch's ANOVA, and Bayesian robust ANOVA are available. Using the R package SSDbain and/or the tables provided in this paper, researchers in the social and behavioral sciences can easily plan the sample size if they intend to use a Bayesian ANOVA.

Bayesian Testing of Scientific Expectations under Multivariate Normal Linear Models

The multivariate normal linear model is one of the most widely employed models for statistical inference in applied research. Special cases include (multivariate) t testing, (M)AN(C)OVA, (multivariate) multiple regression, and repeated measures analysis. Statistical criteria for a model selection problem where models may have equality as well as order constraints on the model parameters based on scientific expectations are limited however. This paper presents a default Bayes factor for this inference problem using fractional Bayes methodology. Group specific fractions are used to properly control prior information. Furthermore the fractional prior is centered on the boundary of the constrained space to properly evaluate order-constrained models. The criterion enjoys various important properties under a broad set of testing problems. The methodology is readily usable via the R package 'BFpack'. Applications from the social and medical sciences are provided to illustrate the methodology.

General Pavlovian-to-instrumental transfer in humans: Evidence from Bayesian inference

When repeatedly paired with rewarding outcomes (i.e., Pavlovian conditioning), environmental cues may acquire predictive and motivational significance and later enhance instrumental responding for the same (i.e., outcome-specific transfer) or motivationally similar (i.e., general transfer) outcomes. Although outcome-specific and general Pavlovian-to-Instrumental Transfer (PIT) are characterized by different neural substrates and behavioral mechanisms, general transfer has never been studied in isolation from outcome-specific transfer in humans. The first aim of the present study was to test whether the general transfer effect could emerge in isolation and independently of outcome-specific transfer. Our results showed that general transfer can be elicited without the concurrent presence of outcome-specific transfer, supporting the idea that outcome-specific and general transfer can be studied independently of each other. The second aim of the present study was to clarify whether the affordance-like properties of the outcomes can affect the general transfer. In fact, a critical difference in current studies on general transfer concerns the use of cues associated with outcomes for which an action was previously learned (or not) during the instrumental training. This apparently minor difference affects the affordance-like properties of the outcome and may also be transferred to the cue, in turn impacting general transfer. Results revealed a general transfer of the same magnitude regardless of whether cues were associated with reward earned or not during instrumental conditioning. These findings increase the current knowledge on the incentive motivational mechanism behind general transfer, indicating that it is independent of the motor features of the outcome.

Dynamic relational event modeling: Testing, exploring, and applying

The relational event model (REM) facilitates the study of network evolution in relational event history data, i.e., time-ordered sequences of social interactions. In real-life social networks it is likely that network effects, i.e., the parameters that quantify the relative importance of drivers of these social interaction sequences, change over time. In these networks, the basic REM is not appropriate to understand what drives network evolution. This research extends the REM framework with approaches for testing and exploring time-varying network effects. First, we develop a Bayesian approach to test whether network effects change during the study period. We conduct a simulation study that illustrates that the Bayesian test accurately quantifies the evidence between a basic (‘static’) REM or a dynamic REM. Secondly, in the case of the latter, time-varying network effects can be studied by means of a moving window that slides over the relational event history. A simulation study was conducted that illustrates that the accuracy and precision of the estimates depend on the window width: narrower windows result in greater accuracy at the cost of lower precision. Third, we develop a Bayesian approach for determining window widths using the empirical network data and conduct a simulation study that illustrates that estimation with empirically determined window widths achieves both good accuracy for time intervals with important changes and good precision for time intervals with hardly any changes in the effects. Finally, in an empirical application, we illustrate how the approaches in this research can be used to test for and explore time-varying network effects of face-to-face contacts at the workplace.

Interaction effect: Are you doing the right thing?

How to correctly interpret interaction effects has been largely discussed in scientific literature. Nevertheless, misinterpretations are still frequently observed, and neuroscience is not exempt from this trend. We reviewed 645 papers published from 2019 to 2020 and found that, in the 93.2% of studies reporting a statistically significant interaction effect (N = 221), post-hoc pairwise comparisons were the designated method adopted to interpret its results. Given the widespread use of this approach, we aim to: (1) highlight its limitations and how it can lead to misinterpretations of the interaction effect; (2) discuss more effective and powerful ways to correctly interpret interaction effects, including both explorative and model selection procedures. The paper provides practical examples and freely accessible online materials to reproduce all analyses.

Empirical Investigation of the Verbal Cues Involved in Delivering Experiential Metaphors

Delivering metaphors experientially has been emphasized in several psychotherapies, such as acceptance and commitment therapy. However, few research has analyzed the variables involved in the efficacy of metaphors. This experimental analog study aims to advance in this topic by analyzing the effect of two components involved in the experiential delivery of metaphors in psychotherapy. The first component is presenting the metaphor by asking the individual to imagine herself as the protagonist of the story versus presenting the metaphor in the third person (Self vs. Other). The second component is the inclusion of verbal cues prompting the relational elaboration of the rules derived from the metaphor content versus not including these prompts (Elaboration vs. No Elaboration). The effect of these components was tested in a double-blind, randomized, 2 × 2 factorial experiment that used the cold pressor task (CPT). Eighty-four participants were exposed to the CPT at the pretest. Afterward, participants were randomly assigned to four experimental protocols. The protocols were audiotaped and consisted of the same metaphor presented in four slightly different ways. Specifically, the protocol of Condition A involved a metaphor with Self and Elaboration, Condition B involved Self and No Elaboration, Condition C involved Other and Elaboration, and Condition D involved Other and No Elaboration. Then, participants were re-exposed to the CPT in the posttest. We hypothesized that Condition A (Self and Elaboration) would show a higher mean increase in pain tolerance than the remaining conditions, which would show similar results. The results were consistent with this hypothesis because Condition A showed a higher percentual increase in pain tolerance (Condition A: M = 268.21, SD = 167.47; Condition B: M = 180.86, SD = 73.01; Condition C: M = 204.81, SD = 100.19; Condition D: M = 175.41, SD = 76.00). A Bayesian informative hypothesis evaluation showed that this hypothesis obtained the highest posterior model probability. Thus, the results indicate that introducing metaphors by asking the individual to imagine herself as the protagonist of the story and providing prompts for relational elaboration might increase the therapeutic effect of the metaphor.

Bayesian updating: increasing sample size during the course of a study

Background

A priori sample size calculation requires an a priori estimate of the size of the effect. An incorrect estimate may result in a sample size that is too low to detect effects or that is unnecessarily high. An alternative to a priori sample size calculation is Bayesian updating, a procedure that allows increasing sample size during the course of a study until sufficient support for a hypothesis is achieved. This procedure does not require and a priori estimate of the effect size. This paper introduces Bayesian updating to researchers in the biomedical field and presents a simulation study that gives insight in sample sizes that may be expected for two-group comparisons.

Methods

Bayesian updating uses the Bayes factor, which quantifies the degree of support for a hypothesis versus another one given the data. It can be re-calculated each time new subjects are added, without the need to correct for multiple interim analyses. A simulation study was conducted to study what sample size may be expected and how large the error rate is, that is, how often the Bayes factor shows most support for the hypothesis that was not used to generate the data.

Results

The results of the simulation study are presented in a Shiny app and summarized in this paper. Lower sample size is expected when the effect size is larger and the required degree of support is lower. However, larger error rates may be observed when a low degree of support is required and/or when the sample size at the start of the study is small. Furthermore, it may occur sufficient support for neither hypothesis is achieved when the sample size is bounded by a maximum.

Conclusions

Bayesian updating is a useful alternative to a priori sample size calculation, especially so in studies where additional subjects can be recruited easily and data become available in a limited amount of time. The results of the simulation study show how large a sample size can be expected and how large the error rate is.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12874-021-01334-6.

Sample-size determination for the Bayesian t test and Welch's test using the approximate adjusted fractional Bayes factor

When two independent means μ₁ and μ₂ are compared, H₀ : μ₁ = μ₂, H₁ : μ₁≠μ₂, and H₂ : μ₁ > μ₂ are the hypotheses of interest. This paper introduces the R package SSDbain, which can be used to determine the sample size needed to evaluate these hypotheses using the approximate adjusted fractional Bayes factor (AAFBF) implemented in the R package bain. Both the Bayesian t test and the Bayesian Welch’s test are available in this R package. The sample size required will be calculated such that the probability that the Bayes factor is larger than a threshold value is at least η if either the null or alternative hypothesis is true. Using the R package SSDbain and/or the tables provided in this paper, psychological researchers can easily determine the required sample size for their experiments.

Perception of action-outcomes is shaped by life-long and contextual expectations

The way humans perceive the outcomes of their actions is strongly colored by their expectations. These expectations can develop over different timescales and are not always complementary. The present work examines how long-term (structural) expectations - developed over a lifetime - and short-term (contextual) expectations jointly affect perception. In two studies, including a pre-registered replication, participants initiated the movement of an ambiguously rotating sphere by operating a rotary switch. In the absence of any learning, participants predominantly perceived the sphere to rotate in the same direction as their rotary action. This bias toward structural expectations was abolished (but not reversed) when participants were exposed to incompatible action-effect contingencies (e.g., clockwise actions causing counterclockwise percepts) during a preceding learning phase. Exposure to compatible action-effect contingencies, however, did not add to the existing structural bias. Together, these findings reveal that perception of action-outcomes results from the combined influence of both long-term and immediate expectations.