Erich H. Witte

In psychology as elsewhere, the main statistical inference strategy to establish empirical effects is null-hypothesis significance testing (NHST). The recent failure to replicate allegedly well-established NHST-results, however, implies that such results lack sufficient statistical power, and thus feature unacceptably high error-rates. Using data-simulation to estimate the error-rates of NHST-results, we advocate the research program strategy (RPS) as a superior methodology. RPS integrates Frequentist with Bayesian inference elements, and leads from a preliminary discovery against a (random) H 0-hypothesis to a statistical H 1-verification. Not only do RPS-results feature significantly lower error-rates than NHST-results, RPS also addresses key-deficits of a " pure " Frequentist and a standard Bayesian approach. In particular, RPS aggregates underpowered results safely. RPS therefore provides a tool to regain the trust the discipline had lost during the ongoing replicability-crisis.

Meiner eigenen Historie entnehme ich, dass ich im 1. Semester (SoSe 1965) eine vierstündige Vorlesung bei Hofstätter mit dem Titel Geschichte und System der Psychologie besucht habe. Aus der Rückschau kann ich sagen, dass ich die Inhalte nur mäßig interessant fand und diese für mich wenig verständlich waren. Letztlich wollte ich als Erstsemester die neue Psychologie und nicht die alte. Als Hofstätter-Schüler war es für mich aber im Laufe des Studiums ganz normal, sich mit den Ursprüngen der Psychologie und ihre Einbettung in kulturelle Entwicklungen zu beschäftigen. Das wurde immer auch mitdiskutiert. Auf diese Weise entwickelte sich ein Bezug zu den aktuellen Themen, die in ihrer Ausdifferenziertheit durch den historischen Bezug eine Verbindung erhalten: Der Wald wird bei den vielen Bäumen sichtbar. Das vermeintlich Neue wird als Variation des Alten interpretierbar, lässt sich einordnen und die unübersichtliche Forschungsvielfalt kann besser verarbeitet und verknüpft werden. Das hilft dem Forschenden. Gewinnt jedoch die aktuelle Forschung auch durch ein Teilfach, das sich intensiv mit der Geschichte beschäftigt? Das ist letztlich die entscheidende Frage. Hierzu möchte ich die Geschichte der Psychologie in folgende fünf Lehr-und Forschungsbereiche untergliedern: 1. Personengeschichte-Wer hat was wann gesagt? 2. Institutionen-Geschichte-Welche Institutionen sind wann gegründet worden und haben sich mit welchen Inhalten beschäftigt? 3. Theorien-und Methodengeschichte-Wie wurden psychologische Inhalte betrachtet, welche Zugangsweisen hat es gegeben und welche Kritikpunkte? 4. Ergebnisgeschichte-Welche relevanten Forschungsergebnisse wurden gefunden und welche Bezüge gibt es zur heutigen Forschung? 5. Epochengeschichte-In welchem Zeitraum wurden welche theoretischen und methodischen Fragen behandelt, welche Umwelteinflüsse (Zeitgeist, Instrumente) gab es und welche Entwicklungen über die Zeit kann man erkennen? Die ersten beiden Punkte gehören zur Identitätsbildung der Forschenden als Psychologen in der Kooperation mit anderen Fächern. Wer kennt nicht die Fragen bei entsprechenden Inhalten: Sind Sie Jurist? Sind Sie Mediziner? Wann fragen wir, Sind Sie Psychologe? Was die nächsten beiden Punkte angeht, so scheint mir nicht selten das Wissen um die Vergangenheit dem Wissen um die Zukunft zu entsprechen. Nur muss das bekanntlich nicht so sein. Man kann nämlich nachschauen. Das ist historische Forschung. Wann man die Geschichte der Inferenzstatistik genauer betrachtet, dann dürfte man über die aktuellen Probleme (keine Replikationsstudien, geringe power etc.) nicht erstaunt sein. Das ist alles umfänglich diskutiert worden (Witte, 1980). Wenn man sich die methodologischen Positionen zur Untersuchung psychologischer Phänomene anschaut, dann werden schon bei Aristoteles vier Arten von Ursachen genannt, die auch heute noch unsere Methodologie differenzieren und integrieren können (Witte, 1987). Unterschiedliche Epochen haben verschiedene Schwerpunktbildungen bei den Ursachen und damit bei den Menschenbildern. Wenn man die heutige psychologische Forschung betrachtet, dann trifft man immer wieder auf die Dreiteilung in kognitiv, affektiv und konativ (behavioral). Ist diese Dreiteilung nur eine Übernahme (ideologisch geprägtes, unproblematisches Hintergrundwissen) von Sokrates oder gibt es wissenschaftlich fundierte Hinweise? Hier dient die Geschichte der kritischen Reflexion mit ihren epochalen Verschiebungen. Wenn wir auf vergessene Ergebnisse schauen, die in der Literatur berichtet werden und für die heutige Forschung von großer Relevanz sein können, so möchte ich den Köhler-Effekt nennen, der als Gegensatz zum Ringelmann-Effekt eine Steigerung der Gruppenleistung zur Folge hat, aber völlig ignoriert worden ist (Witte,1989). Heute ist er ein Standard-Effekt.

ABSTRACT
We reconstruct recent work on macrosocial stress (Chou, Parmar, & Galinsky, 2016) as if it were an instance of a research strategy that tests point-alternative hypotheses within a full-fledged Research program. Because this strategy is free of various deficits that beset dominant strategies (e.g., metaanalysis, Bayes factor analysis), our article demonstrates one way in which the confidence crisis may be overcome.

Zusammenfassung. Die vorliegende Studie untersucht die disziplinäre Matrix (vgl. Kuhn, 1974,1978) der Psychologie im Vergleich zur Physik, Biologie und Sozialwissenschaft in Hinsicht auf die Forschungsmethoden in diesen Disziplinen. 1200 Forschungsleiter und Mitarbeiter von DFG-geförderten Projekten wurden gebeten einen Online-Fragebogen zu ihrer Forschungsmethodik auszufüllen. Die resultierende Gesamtstichprobe umfasst n = 594 Projekte. Eine grundsätzliche Vergleichbarkeit der Disziplinen zeigt sich in den Kor-relationen ihrer Korrelationsmatrizen und in der Faktorenanalyse. Eine einfaktorielle Varianzanalyse ergab methodologische Unter-schiede zwischen den Fächern, die sich bei dem Vergleich von prototypischen Projekten noch deutlicher gestalteten. Akademische Prototypen sind hier solche Projekte, die einem Fach angehören und gleichzeitig dem Cluster, das durch eine Mehrheit desselben Fachs bestimmt ist. Für die Psychologie zeigen die Ergebnisse eine deutlichere Orientierung am quantitativ-hypothesentestenden Ansatz als in allen anderen Fächern. Die prototypische Psychologie konzentriert sich außerdem stärker auf das Testen von Theorien als alle anderen Fächer. Schlüsselwörter: disziplinäre Matrix, Methodologie, Forschungsmethoden, Wissenschaftstheorie Research in psychology. Its disciplinary matrix as compared to physics, biology, and social science Abstract. The present study explores the disciplinary matrix of psychology (see Kuhn, 1974, 1978) in terms of research methods and compares it to the matrices of physics, biology, and social sciences. Research directors and staff members of 1200 research projects funded by the German Research Foundation (DFG) were asked to complete an online questionnaire, resulting in a final sample of n = 594 projects. A general comparability of the four disciplines is demonstrated with correlations of their correlation matrices and factor analysis. ANOVA detects differences between the methods of the academic disciplines' which become more apparent when only the prototypical projects within each discipline are analyzed. These academic prototypes are projects of a discipline belonging to the methods cluster that is defined by a majority of projects of that discipline. Results indicate that psychology follows the quantitative, hypothesis-testing approach significantly more than all other disciplines. The psychological prototype also focuses significantly more on testing theories than the other disciplines.

In this study, contemporary creativity research was investigated by conducting an analysis of dissertation abstracts. Keywords relating to creativity were used to collect a sample from an online database. The abstracts were coded by discipline, research aspect (trait, process, or product), social level of analysis (individual, group, organization, or culture), and research approach (empirical vs. theoretical; quantitative vs. qualitative). Classification

Two large online surveys were conducted among employees in Germany to explore the importance employees and organizations lay on aspects of interpersonal respect in relation to other work values. The first study (N1 = 589) extracted a general ranking of work values, showing ...

The effect size measure narrowly concerns the causal influence, or correlation, between an experimental setting's independent and dependent variables. A given empirical effect's significance, by contrast, pivots on its statistical, theoretical, and practical aspect-that is, the effect's magnitude, how well a theory models it, and the utility we ascribe to it. Rather than mix these aspects, a sound evaluation must keep them apart. Once we evaluate theories by data, however, effect size measures that model participants' average reaction cannot stand proxy for participants' observed individual reaction. To facilitate the construction of valid, informative theories of intended phenomena, we present and exemplify three indices that assess a theory's quality in terms of its data similarity, reliability, and its validity.

To justify the effort of developing a theoretical construct, a theoretician needs data that bear out a non-random effect of sufficiently high replication probability. To establish such effects statistically, researchers (rightly) rely on a t-test. But many pursue questionable strategies meant to lower the cost of data collection. Our paper reconstructs two such strategies. Both reduce the minimum sample-size (NMIN) necessary under conventional error rates (α, β) to register a given effect size (d) as statistically significant. The first strategy increases the β-error rate; the second treats the control group as a constant, thereby collapsing a two sample t-test into its one sample version. As an example, a two sample t-test for d=0.50 given α=β=0.05 requiring NMIN=176 becomes a one sample t-test given α=0.05, β=0.20 requiring NMIN=27. Not only does this decrease the replication probability from (1-β)=0.95 to (1-β)=0.80, but worse, the second strategy collapses a Neyman Pearson-test into a Fisher-test, which cannot corroborate hypotheses meaningfully. The ubiquity of both strategies arguably makes them partial causes of the confidence crisis. But since research groups can collaborate to reach NMIN jointly, their individual resource limitations justify neither strategy.

The effect size measures the causal influence, or correlation, between an experimental setting’s independent and dependent variable. But an effect’s overall significance pivots on its statistical, theoretical, and practical aspect, that is, the effect’s magnitude, how well a theory predicts it, and the utility ascribed to it. Rather than mix these three aspects, a sound evaluation of empirical effects keeps them apart. When evaluating theories by data, moreover, effect sizes for participants’ theoretically modelled average reaction cannot stand proxy for their observed individual reactions. To facilitate constructing valid, informative theories of an intended phenomenon, we present and exemplify two indices that assess a theory’s quality as its data similarity and its validity.

To justify the effort of developing a theoretical construct, a theoretician needs empirical data that support a non-random effect of sufficiently high replication-probability. To establish these effects statistically, researchers (rightly) rely on a t-test. But many pursue questionable strategies that lower the cost of data-collection. Our paper reconstructs two such strategies. Both reduce the minimum sample-size (NMIN) sufficing under conventional errors (α, β) to register a given effect-size (d) as a statistically significant non-random data signature. The first strategy increases the β-error; the second treats the control-group as a constant, thereby collapsing a two-sample t-test into its one-sample version. (A two-sample t-test for d=0.50 under α=β=0.05 with NMIN=176, for instance, becomes a one-sample t-test under α=0.05, β=0.20 with NMIN=27.) Not only does this decrease the replication-probability of data from (1-β)=0.95 to (1-β)=0.80, particularly the second strategy cannot corroborate hypotheses meaningfully. The ubiquity of both strategies arguably makes them partial causes of the confidence-crisis. But as resource-pooling would allow research groups to reach NMIN jointly, a group’s individually limited resources justify neither strategy.

The gold standard for an empirical science is the replicability of its research results. But the estimated average replicability rate of key-effects that top-tier psychology journals report falls between 36 and 39% (objective vs. subjective rate; Open Science Collaboration, 2015). So the standard mode of applying null-hypothesis significance testing (NHST) fails to adequately separate stable from random effects. Therefore, NHST does not fully convince as a statistical inference strategy. We argue that the replicability crisis is “home-made” because more sophisticated strategies can deliver results the successful replication of which is sufficiently probable. Thus, we can overcome the replicability crisis by integrating empirical results into genuine research programs. Instead of continuing to narrowly evaluate only the stability of data against random fluctuations (discovery context), such programs evaluate rival hypotheses against stable data (justification context).

The main intentions of this paper are to discuss the
similarities and dissimilarities between different dynamic
models of social influence in small groups and the development
of a dynamic version of the group situation theory.
As theoretical approaches were chosen : a) the social transition
scheme model (STS) developed by Kerr(1981, 1982)  b) the social interaction  sequence model (SlS) proposed by Stasser & Davis(1981), and c) the dynanic theory of social irnpact (DTSI)
pubtished by Nowak' Szamrej and Latane'(1990). These theories
were compared with group situation theory developed by Witte (1987,90) and now modified as a dynanic version. This group situation theory tried to explain the different meanings of a group decision for their members after the group discussion and not only the change of the opinions to reach a consensus. However, this quatitative change of the Group situation in its normative components has to be modelled in the
future.

Group situation theory has been developed as a general-framework in order to integrate different theoretical approacbes in the area of small group research. Its development is based on some fundamental assumptions combining normative influences with an information integration process of the individual group member. Under these basic assumptions as guidelines the social impact theory has been integrated into the group situation theory which results in an extended group situation theory.
After this extension the similarities of this approach with the social decision schemes has been discussed. The predictions are very similar, but the kind of theories are different: Social decision schemes is a family of descriptive models predicting the transformation of a group members’ distribution after discussion using qualitative choices; EGST is a family of information integration processes giving an explanation why a given change has happened after discussion using also quantitative choices.
As a next step the BALES’ research tradition on interaction frequencies and their influence in group decisions has been integrated in EGST to find their specific place in a broader theoretical concept.

Published in Frontiers in Psychology, at https://www.frontiersin.org/articles/10.3389/fpsyg.2018.00460/

In psychology as elsewhere, the main statistical inference strategy to establish empirical effects is null-hypothesis significance testing (NHST). The recent failure to replicate allegedly well-established NHST-results, however, implies that such results lack sufficient statistical power, and thus feature unacceptably high error-rates. Using data-simulation to estimate the error-rates of NHST-results, we advocate the research program strategy (RPS) as a superior methodology. RPS integrates Frequentist with Bayesian inference elements, and leads from a preliminary discovery against a (random) H0-hypothesis to a statistical H1-verification. Not only do RPS-results feature significantly lower error-rates than NHST-results, RPS also addresses key-deficits of a “pure” Frequentist and a standard Bayesian approach. In particular, RPS aggregates underpowered results safely. RPS therefore provides a tool to regain the trust the discipline had lost in the ongoing replicability-crisis.

Commentary on: Rolf A. Zwaan, Alexander Etz, Richard E. Lucas, and M. Brent Donnellan (2017). Making Replication Mainstream. Behavioral and Brain Sciences, published online: 25 October 2017, pp. 1-50; forthcoming at https://www.cambridge.org/core/journals/behavioral-and-brain-sciences *******Abstract:
Before replication becomes mainstream, the potential for generating theoretical knowledge better be clear. Replicating statistically significant non-random data shows that an original study made a discovery; replicating a specified theoretical effect shows that an original study corroborated a theory. Yet only in the latter case is replication a necessary, sound, and worthwhile strategy.

We reanalyze the recent multilab preregistered study on ego-depletion by Hagger and Chatzisarantis (2016) as if their data were obtained under the research program-strategy (Witte & Zenker, 2016a, 2016b). This strengthens Hagger and Chatzisarantis’s (2016) main conclusion, because our reanalysis more directly corroborates the absence of a medium-sized, or a small-sized, ego-depletion effect (d ¼ .50 under α ¼ β ¼ .05; d ¼ .20 under α ¼ β ¼ .01). We explain how a smaller ego-depletion effect of d ¼ .04 can be tested under similar conditions, having determined this value by maximum likelihood estimation, and compare the research program-strategy to a standard meta-analytic integration.

We reconstruct recent work on macro-social stress (Chou et al., 2016) as if it were an instance of a research strategy that tests point-alternative hypotheses within a full-fledged research program. Since this strategy is free of various deficits that beset dominant strategies (e.g., meta-analysis, Bayes-factor analysis), our article demonstrates one way in which the confidence crisis may be overcome.