Same data, different conclusions: Radical dispersion in empirical results when independent analysts operationalize and test the same hypothesis

Toward a more credible assessment of the credibility of science by many-analyst studies

We discuss a relatively new meta-scientific research design: many-analyst studies that attempt to assess the replicability and credibility of research based on large-scale observational data. In these studies, a large number of analysts try to answer the same research question using the same data. The key idea is the greater the variation in results, the greater the uncertainty in answering the research question and, accordingly, the lower the credibility of any individual research finding. Compared to individual replications, the large crowd of analysts allows for a more systematic investigation of uncertainty and its sources. However, many-analyst studies are also resource-intensive, and there are some doubts about their potential to provide credible assessments. We identify three issues that any many-analyst study must address: 1) identifying the source of variation in the results; 2) providing an incentive structure similar to that of standard research; and 3) conducting a proper meta-analysis of the results. We argue that some recent many-analyst studies have failed to address these issues satisfactorily and have therefore provided an overly pessimistic assessment of the credibility of science. We also provide some concrete guidance on how future many-analyst studies could provide a more constructive assessment.

Prevalence, patterns and predictors of paranormal beliefs in The Netherlands: a several-analysts approach

Paranormal beliefs encompass a wide variety of phenomena, including the existence of supernatural entities such as ghosts and witches, as well as extraordinary human abilities such as telepathy and clairvoyance. In the current study, we used a nationally representative sample ( N = 2534 ) to investigate the presence and correlates of paranormal beliefs among the secular Dutch population. The results indicated that most single paranormal phenomena (e.g. belief in clairvoyance) are endorsed by 10-20% of Dutch respondents; however, 55.6% of respondents qualify as paranormal believers based on the preregistered criterion that they believe in at least one phenomenon with considerable certainty. In addition, we invited four analysis teams with different methodological expertise to assess the structure of paranormal beliefs using traditional factor analysis, network analysis, Bayesian network analysis and latent class analysis (LCA). The teams' analyses indicated adequate fit of a four-factor structure reported in a 1985 study, but also emphasized different conclusions across techniques; network analyses showed evidence against strong connectedness within most clusters, and suggested a five-cluster structure. The application of various analytic techniques painted a nuanced picture of paranormal beliefs and believers in The Netherlands and suggests that despite increased secularization, subgroups of the general population still believe in paranormal phenomena.

Heterogeneity in effect size estimates

A typical empirical study involves choosing a sample, a research design, and an analysis path. Variation in such choices across studies leads to heterogeneity in results that introduce an additional layer of uncertainty, limiting the generalizability of published scientific findings. We provide a framework for studying heterogeneity in the social sciences and divide heterogeneity into population, design, and analytical heterogeneity. Our framework suggests that after accounting for heterogeneity, the probability that the tested hypothesis is true for the average population, design, and analysis path can be much lower than implied by nominal error rates of statistically significant individual studies. We estimate each type's heterogeneity from 70 multilab replication studies, 11 prospective meta-analyses of studies employing different experimental designs, and 5 multianalyst studies. In our data, population heterogeneity tends to be relatively small, whereas design and analytical heterogeneity are large. Our results should, however, be interpreted cautiously due to the limited number of studies and the large uncertainty in the heterogeneity estimates. We discuss several ways to parse and account for heterogeneity in the context of different methodologies.

The multiverse of universes: A tutorial to plan, execute and interpret multiverses analyses using the R package multiverse

Even when guided by strong theories and sound methods, researchers must often choose a singular course of action from multiple viable alternatives. Regardless of the choice, it, along with all other choices made during the research process, individually and collectively affects study results, often in unpredictable ways. The inability to disentangle how much of an observed effect is attributable to the phenomenon of interest, and how much is attributable to what have come to be known as researcher degrees of freedom (RDF), slows theoretical progress and stymies practical implementation. However, if one could examine the results from a particular set of RDF (known as a universe) against a systematically and comprehensively determined background of alternative viable universes (known as a multiverse), then the effects of RDF can be directly examined to provide greater context and clarity to future researchers, and greater confidence in the recommendations to practitioners. This tutorial demonstrates a means to map result variability directly and efficiently, and empirically investigate RDF impact on conclusions via multiverse analysis. Using the R package multiverse, we outline best practices in planning, executing and interpreting of multiverse analyses.

Subjective evidence evaluation survey for many-analysts studies

Many-analysts studies explore how well an empirical claim withstands plausible alternative analyses of the same dataset by multiple, independent analysis teams. Conclusions from these studies typically rely on a single outcome metric (e.g. effect size) provided by each analysis team. Although informative about the range of plausible effects in a dataset, a single effect size from each team does not provide a complete, nuanced understanding of how analysis choices are related to the outcome. We used the Delphi consensus technique with input from 37 experts to develop an 18-item subjective evidence evaluation survey (SEES) to evaluate how each analysis team views the methodological appropriateness of the research design and the strength of evidence for the hypothesis. We illustrate the usefulness of the SEES in providing richer evidence assessment with pilot data from a previous many-analysts study.

Prediction Challenge: Simulating Rydberg photoexcited cyclobutanone with surface hopping dynamics based on different electronic structure methods

This research examines the nonadiabatic dynamics of cyclobutanone after excitation into the n → 3s Rydberg S2 state. It stems from our contribution to the Special Topic of the Journal of Chemical Physics to test the predictive capability of computational chemistry against unseen experimental data. Decoherence-corrected fewest-switches surface hopping was used to simulate nonadiabatic dynamics with full and approximated nonadiabatic couplings. Several simulation sets were computed with different electronic structure methods, including a multiconfigurational wavefunction [multiconfigurational self-consistent field (MCSCF)] specially built to describe dissociative channels, multireference semiempirical approach, time-dependent density functional theory, algebraic diagrammatic construction, and coupled cluster. MCSCF dynamics predicts a slow deactivation of the S2 state (10 ps), followed by an ultrafast population transfer from S1 to S0 (<100 fs). CO elimination (C3 channel) dominates over C2H4 formation (C2 channel). These findings radically differ from the other methods, which predicted S2 lifetimes 10-250 times shorter and C2 channel predominance. These results suggest that routine electronic structure methods may hold low predictive power for the outcome of nonadiabatic dynamics.

We urgently need a culture of multi-operationalization in psychological research

Analysis of different operationalizations shows that many scientific results may be an artifact of the operationalization process. A culture of multi-operationalization may be needed for psychological research to develop valid knowledge.

Gender Bias in Letters of Recommendation: Relevance to Urology Match Outcomes and Pursuit of Fellowship Training/Academic Career

OBJECTIVE

To review applications to a single urology residency program to determine application characteristics predictive of (1) successful match into urology residency and (2) pursuit of fellowship training and/or academic practice after completion of residency. Our principal variables of interest were gender bias as assessed in letters of recommendation (LOR), personal statements, Medical Student Performance Evaluation (MSPE), race, and gender.

MATERIALS AND METHODS

Applications submitted to our urology residency program in the 2014 cycle were reviewed. Twenty-three variables were analyzed, including applicant demographics, application materials, and gender bias. Deidentified text from LOR, personal statements, and MSPE was evaluated for gender bias using an open-source gender bias calculator. A subanalysis of applicants who matched at a top 25 urology program was performed. Logistic regression analysis was performed to identify applicant variables associated with (1) match success and (2) fellowship training or academic employment as of September 2021.

RESULTS

Two hundred and twenty-two completed applications were analyzed. First authorship of a published manuscript was significantly associated with greater odds of matching. Female gender and top 25 medical school attendance were both significant predictors of matching at a top 25 urology program. The number of first-author publications was associated with completion of fellowship training or current employment in an academic position.

CONCLUSION

First-author publications are the most important preinterview determinant of match success and subsequent pursuit of academic practice/fellowship training. Certain applicant characteristics are associated with matching at highly ranked programs. Gender bias in application materials (including LOR) does not appear to exert a significant influence on match and early career outcomes.

Informing Harmonization Decisions in Integrative Data Analysis: Exploring the Measurement Multiverse

Combining datasets in an integrative data analysis (IDA) requires researchers to make a number of decisions about how best to harmonize item responses across datasets. This entails two sets of steps: logical harmonization, which involves combining items which appear similar across datasets, and analytic harmonization, which involves using psychometric models to find and account for cross-study differences in measurement. Embedded in logical and analytic harmonization are many decisions, from deciding whether items can be combined prima facie to how best to find covariate effects on specific items. Researchers may not have specific hypotheses about these decisions, and each individual choice may seem arbitrary, but the cumulative effects of these decisions are unknown. In the current study, we conducted an IDA of the relationship between alcohol use and delinquency using three datasets (total N = 2245). For analytic harmonization, we used moderated nonlinear factor analysis (MNLFA) to generate factor scores for delinquency. We conducted both logical and analytic harmonization 72 times, each time making a different set of decisions. We assessed the cumulative influence of these decisions on MNLFA parameter estimates, factor scores, and estimates of the relationship between delinquency and alcohol use. There were differences across paths in MNLFA parameter estimates, but fewer differences in estimates of factor scores and regression parameters linking delinquency to alcohol use. These results suggest that factor scores may be relatively robust to subtly different decisions in data harmonization, and measurement model parameters are less so.

The pupil collaboration: A multi-lab, multi-method analysis of goal attribution in infants

The rise of pupillometry in infant research over the last decade is associated with a variety of methods for data preprocessing and analysis. Although pupil diameter is increasingly recognized as an alternative measure of the popular cumulative looking time approach used in many studies (Jackson & Sirois, 2022), an open question is whether the many approaches used to analyse this variable converge. To this end, we proposed a crowdsourced approach to pupillometry analysis. A dataset from 30 9-month-old infants (15 girls; Mage = 282.9 days, SD = 8.10) was provided to 7 distinct teams for analysis. The data were obtained from infants watching video sequences showing a hand, initially resting between two toys, grabbing one of them (after Woodward, 1998). After habituation, infants were shown (in random order) a sequence of four test events that varied target position and target toy. Results show that looking times reflect primarily the familiar path of the hand, regardless of target toy. Gaze data similarly show this familiarity effect of path. The pupil dilation analyses show that features of pupil baseline measures (duration and temporal location) as well as data retention variation (trial and/or participant) due to different inclusion criteria from the various analysis methods are linked to divergences in findings. Two of the seven teams found no significant findings, whereas the remaining five teams differ in the pattern of findings for main and interaction effects. The discussion proposes guidelines for best practice in the analysis of pupillometry data.

The positive-negative-competence (PNC) model of psychological responses to representations of robots

Robots are becoming an increasingly prominent part of society. Despite their growing importance, there exists no overarching model that synthesizes people's psychological reactions to robots and identifies what factors shape them. To address this, we created a taxonomy of affective, cognitive and behavioural processes in response to a comprehensive stimulus sample depicting robots from 28 domains of human activity (for example, education, hospitality and industry) and examined its individual difference predictors. Across seven studies that tested 9,274 UK and US participants recruited via online panels, we used a data-driven approach combining qualitative and quantitative techniques to develop the positive-negative-competence model, which categorizes all psychological processes in response to the stimulus sample into three dimensions: positive, negative and competence-related. We also established the main individual difference predictors of these dimensions and examined the mechanisms for each predictor. Overall, this research provides an in-depth understanding of psychological functioning regarding representations of robots.

Research-Problem Validity in Primary Research: Precision and Transparency in Characterizing Past Knowledge

Four validity types evaluate the approximate truth of inferences communicated by primary research. However, current validity frameworks ignore the truthfulness of empirical inferences that are central to research-problem statements. Problem statements contrast a review of past research with other knowledge that extends, contradicts, or calls into question specific features of past research. Authors communicate empirical inferences, or quantitative judgments, about the frequency (e.g., "few," "most") and variability (e.g., "on the one hand," "on the other hand") in their reviews of existing theories, measures, samples, or results. We code a random sample of primary research articles and show that 83% of quantitative judgments in our sample are vague and do not have a transparent origin, making it difficult to assess their validity. We review validity threats of current practices. We propose that documenting the literature search, reporting how the search was coded, and quantifying the search results facilitates more precise judgments and makes their origin transparent. This practice enables research questions that are more closely tied to the existing body of knowledge and allows for more informed evaluations of the contribution of primary research articles, their design choices, and how they advance knowledge. We discuss potential limitations of our proposed framework.

The trustworthiness of the cumulative knowledge in industrial/organizational psychology: The current state of affairs and a path forward

The goal of industrial/organizational (IO) psychology, is to build and organize trustworthy knowledge about people-related phenomena in the workplace. Unfortunately, as with other scientific disciplines, our discipline may be experiencing a "crisis of confidence" stemming from the lack of reproducibility and replicability of many of our field's research findings, which would suggest that much of our research may be untrustworthy. If a scientific discipline's research is deemed untrustworthy, it can have dire consequences, including the withdraw of funding for future research. In this focal article, we review the current state of reproducibility and replicability in IO psychology and related fields. As part of this review, we discuss factors that make it less likely that research findings will be trustworthy, including the prevalence of scientific misconduct, questionable research practices (QRPs), and errors. We then identify some root causes of these issues and provide several potential remedies. In particular, we highlight the need for improved research methods and statistics training as well as a re-alignment of the incentive structure in academia. To accomplish this, we advocate for changes in the reward structure, improvements to the peer review process, and the implementation of open science practices. Overall, addressing the current "crisis of confidence" in IO psychology requires individual researchers, academic institutions, and publishers to embrace system-wide change.

Reproducibility in Neuroimaging Analysis: Challenges and Solutions

Recent years have marked a renaissance in efforts to increase research reproducibility in psychology, neuroscience, and related fields. Reproducibility is the cornerstone of a solid foundation of fundamental research-one that will support new theories built on valid findings and technological innovation that works. The increased focus on reproducibility has made the barriers to it increasingly apparent, along with the development of new tools and practices to overcome these barriers. Here, we review challenges, solutions, and emerging best practices with a particular emphasis on neuroimaging studies. We distinguish 3 main types of reproducibility, discussing each in turn. Analytical reproducibility is the ability to reproduce findings using the same data and methods. Replicability is the ability to find an effect in new datasets, using the same or similar methods. Finally, robustness to analytical variability refers to the ability to identify a finding consistently across variation in methods. The incorporation of these tools and practices will result in more reproducible, replicable, and robust psychological and brain research and a stronger scientific foundation across fields of inquiry.

Multiverse analyses of fear acquisition and extinction retention in posttraumatic stress disorder

Persistent fear is a cardinal feature of posttraumatic stress disorder (PTSD), and deficient fear extinction retention is a proposed illness mechanism and target of exposure-based therapy. However, evidence for deficient fear extinction in PTSD has been mixed using laboratory paradigms, which may relate to underidentified methodological variation across studies. We reviewed the literature to identify parameters that differ across studies of fear extinction retention in PTSD. We then performed Multiverse Analysis in a new sample, to quantify the impact of those methodological parameters on statistical findings. In 25 PTSD patients (15 female) and 36 trauma-exposed non-PTSD controls (TENC) (20 female), we recorded skin conductance response (SCR) during fear acquisition and extinction learning (day 1) and extinction recall (day 2). A first Multiverse Analysis examined the effects of methodological parameters identified by the literature review on comparisons of SCR-based fear extinction retention in PTSD versus TENC. A second Multiverse Analysis examined the effects of those methodological parameters on comparisons of SCR to a danger cue (CS+) versus safety cue (CS-) during fear acquisition. Both the literature review and the Multiverse Analysis yielded inconsistent findings for fear extinction retention in PTSD versus TENC, and most analyses found no statistically significant group difference. By contrast, significantly elevated SCR to CS+ versus CS- was consistently found across all analyses in the literature review and the Multiverse Analysis of new data. We discuss methodological parameters that may most contribute to inconsistent findings of fear extinction retention deficit in PTSD and implications for future clinical research.

Preregistration: Practical Considerations for Speech, Language, and Hearing Research

PURPOSE

In the last decade, psychology and other sciences have implemented numerous reforms to improve the robustness of our research, many of which are based on increasing transparency throughout the research process. Among these reforms is the practice of preregistration, in which researchers create a time-stamped and uneditable document before data collection that describes the methods of the study, how the data will be analyzed, the sample size, and many other decisions. The current article highlights the benefits of preregistration with a focus on the specific issues that speech, language, and hearing researchers are likely to encounter, and additionally provides a tutorial for writing preregistrations.

CONCLUSIONS

Although rates of preregistration have increased dramatically in recent years, the practice is still relatively uncommon in research on speech, language, and hearing. Low rates of adoption may be driven by a lack of understanding of the benefits of preregistration (either generally or for our discipline in particular) or uncertainty about how to proceed if it becomes necessary to deviate from the preregistered plan. Alternatively, researchers may see the benefits of preregistration but not know where to start, and gathering this information from a wide variety of sources is arduous and time consuming. This tutorial addresses each of these potential roadblocks to preregistration and equips readers with tools to facilitate writing preregistrations for research on speech, language, and hearing.

SUPPLEMENTAL MATERIAL

https://doi.org/10.23641/asha.21644843.

Examining individual differences in personality trait changes after negative life events

Personality traits can change throughout the entire life span, but people differ in their personality trait changes. To better understand individual differences in personality changes, we examined personal (personality functioning), environmental (environmental changes), and event-related moderators (e.g., perceived event characteristics) of personality trait changes. Therefore, we used a sample of 1069 participants who experienced a negative life event in the last 5 weeks and assessed their personality traits at five measurement occasions over 6 months. Employing preregistered multilevel lasso estimation, we did not find any significant effects. While exploratory analyses generally confirmed this conclusion, they also identified some effects that might being worth to be considered in future research (e.g., perceived impact and perceived social status changes were associated with changes in agreeableness after experiencing a relationship breakup). In total, our moderators explained less than 2% of variance in personality traits. Nonetheless, our study has several important implications for future research on individual differences in personality change. For example, future research should consider personal, environmental, and event-related moderators, use different analytical methods, and rely on highly powered samples to detect very small effects.

Observing many researchers using the same data and hypothesis reveals a hidden universe of uncertainty

This study explores how researchers' analytical choices affect the reliability of scientific findings. Most discussions of reliability problems in science focus on systematic biases. We broaden the lens to emphasize the idiosyncrasy of conscious and unconscious decisions that researchers make during data analysis. We coordinated 161 researchers in 73 research teams and observed their research decisions as they used the same data to independently test the same prominent social science hypothesis: that greater immigration reduces support for social policies among the public. In this typical case of social science research, research teams reported both widely diverging numerical findings and substantive conclusions despite identical start conditions. Researchers' expertise, prior beliefs, and expectations barely predict the wide variation in research outcomes. More than 95% of the total variance in numerical results remains unexplained even after qualitative coding of all identifiable decisions in each team's workflow. This reveals a universe of uncertainty that remains hidden when considering a single study in isolation. The idiosyncratic nature of how researchers' results and conclusions varied is a previously underappreciated explanation for why many scientific hypotheses remain contested. These results call for greater epistemic humility and clarity in reporting scientific findings.

Examining the generalizability of research findings from archival data

The extent to which results from complex datasets generalize across contexts is critically important to numerous scientific fields as well as to practitioners who rely on such analyses to guide important strategic decisions. Our initiative systematically investigated whether findings from the field of strategic management would emerge in new time periods and new geographies. Original findings that were statistically reliable in the first place were typically obtained again in novel tests, suggesting surprisingly little sensitivity to context. For some social scientific areas of inquiry, results from a specific time and place can be a meaningful guide as to what will be observed more generally.

This initiative examined systematically the extent to which a large set of archival research findings generalizes across contexts. We repeated the key analyses for 29 original strategic management effects in the same context (direct reproduction) as well as in 52 novel time periods and geographies; 45% of the reproductions returned results matching the original reports together with 55% of tests in different spans of years and 40% of tests in novel geographies. Some original findings were associated with multiple new tests. Reproducibility was the best predictor of generalizability—for the findings that proved directly reproducible, 84% emerged in other available time periods and 57% emerged in other geographies. Overall, only limited empirical evidence emerged for context sensitivity. In a forecasting survey, independent scientists were able to anticipate which effects would find support in tests in new samples.

Future Objectivity Requires Perspective and Forward Combinatorial Meta-Analyses

This manuscript contributes to a future definition of objectivity by bringing together recent statements in epistemology and methodology. It outlines how improved objectivity can be achieved by systematically incorporating multiple perspectives, thereby improving the validity of science. The more result-biasing perspectives are known, the more a phenomenon of interest can be disentangled from these perspectives. Approaches that call for the integration of perspective into objectivity at the epistemological level or that systematically incorporate different perspectives at the statistical level already exist and are brought together in the manuscript. Recent developments in research methodology, such as transparency, reproducibility of research processes, pre-registration of studies, or free access to raw data, analysis strategies, and syntax, promote the explication of perspectives because they make the entire research process visible. How the explication of perspectives can be done practically is outlined in the manuscript. As a result, future research programs can be organized in such a way that meta-analyses and meta-meta-analyses can be conducted not only backward but forward and prospectively as a regular and thus well-prepared part of objectification and validation processes.

ASAS-NANP symposium: mathematical modeling in animal nutrition: limitations and potential next steps for modeling and modelers in the animal sciences

The field of animal science, and especially animal nutrition, relies heavily on modeling to accomplish its day-to-day objectives. New data streams ("big data") and the exponential increase in computing power have allowed the appearance of "new" modeling methodologies, under the umbrella of artificial intelligence (AI). However, many of these modeling methodologies have been around for decades. According to Gartner, technological innovation follows five distinct phases: technology trigger, peak of inflated expectations, trough of disillusionment, slope of enlightenment, and plateau of productivity. The appearance of AI certainly elicited much hype within agriculture leading to overpromised plug-and-play solutions in a field heavily dependent on custom solutions. The threat of failure can become real when advertising a disruptive innovation as sustainable. This does not mean that we need to abandon AI models. What is most necessary is to demystify the field and place a lesser emphasis on the technology and more on business application. As AI becomes increasingly more powerful and applications start to diverge, new research fields are introduced, and opportunities arise to combine "old" and "new" modeling technologies into hybrids. However, sustainable application is still many years away, and companies and universities alike do well to remain at the forefront. This requires investment in hardware, software, and analytical talent. It also requires a strong connection to the outside world to test, that which does, and does not work in practice and a close view of when the field of agriculture is ready to take its next big steps. Other research fields, such as engineering and automotive, have shown that the application power of AI can be far reaching but only if a realistic view of models as whole is maintained. In this review, we share our view on the current and future limitations of modeling and potential next steps for modelers in the animal sciences. First, we discuss the inherent dependencies and limitations of modeling as a human process. Then, we highlight how models, fueled by AI, can play an enhanced sustainable role in the animal sciences ecosystem. Lastly, we provide recommendations for future animal scientists on how to support themselves, the farmers, and their field, considering the opportunities and challenges the technological innovation brings.

Exposing and overcoming the fixed-effect fallacy through crowd science

By organizing crowds of scientists to independently tackle the same research questions, we can collectively overcome the generalizability crisis. Strategies to draw inferences from a heterogeneous set of research approaches include aggregation, for instance, meta-analyzing the effect sizes obtained by different investigators, and parsing, attempting to identify theoretically meaningful moderators that explain the variability in results.

Reproducibility of Research During COVID-19: Examining the Case of Population Density and the Basic Reproductive Rate from the Perspective of Spatial Analysis

The emergence of the novel SARS-CoV-2 coronavirus and the global COVID-19 pandemic in 2019 led to explosive growth in scientific research. Alas, much of the research in the literature lacks conditions to be reproducible, and recent publications on the association between population density and the basic reproductive number of SARS-CoV-2 are no exception. Relatively few papers share code and data sufficiently, which hinders not only verification but additional experimentation. In this article, an example of reproducible research shows the potential of spatial analysis for epidemiology research during COVID-19. Transparency and openness means that independent researchers can, with only modest efforts, verify findings and use different approaches as appropriate. Given the high stakes of the situation, it is essential that scientific findings, on which good policy depends, are as robust as possible; as the empirical example shows, reproducibility is one of the keys to ensure this.

Consensus-based guidance for conducting and reporting multi-analyst studies

Any large dataset can be analyzed in a number of ways, and it is possible that the use of different analysis strategies will lead to different results and conclusions. One way to assess whether the results obtained depend on the analysis strategy chosen is to employ multiple analysts and leave each of them free to follow their own approach. Here, we present consensus-based guidance for conducting and reporting such multi-analyst studies, and we discuss how broader adoption of the multi-analyst approach has the potential to strengthen the robustness of results and conclusions obtained from analyses of datasets in basic and applied research.

Validation of Machine Learning Models for Structural Dam Behaviour Interpretation and Prediction

The main aim of structural safety control is the multiple assessments of the expected dam behaviour based on models and the measurements and parameters that characterise the dam’s response and condition. In recent years, there is an increase in the use of data-based models for the analysis and interpretation of the structural behaviour of dams. Multiple Linear Regression is the conventional, widely used approach in dam engineering, although interesting results have been published based on machine learning algorithms such as artificial neural networks, support vector machines, random forest, and boosted regression trees. However, these models need to be carefully developed and properly assessed before their application in practice. This is even more relevant when an increase in users of machine learning models is expected. For this reason, this paper presents extensive work regarding the verification and validation of data-based models for the analysis and interpretation of observed dam’s behaviour. This is presented by means of the development of several machine learning models to interpret horizontal displacements in an arch dam in operation. Several validation techniques are applied, including historical data validation, sensitivity analysis, and predictive validation. The results are discussed and conclusions are drawn regarding the practical application of data-based models.