Sven Hilbert

This article investigates whether gold-standard items are helpful for questionnaire development. Based on an initial item pool (138 items) derived through a deductive item construction (𝑁 = 124) and a prototype approach (𝑁 = 24), followed by cognitive interviews (𝑁 = 8) and a pilot study (𝑁 = 390) of a preliminary item set (61 items), three 12-item scales were constructed to measure the physical traits of body height, body weight, and age. We collected data on these scales using response formats with either two (𝑁 = 921) or six (𝑁 = 933) categories. We also collected numeric self-reports of body weight, body height, and age as gold-standard items. We argue that the items intended to measure the latent variable (i.e., body height, body weight, age) most directly (i.e., the gold-standard items) should have the highest loadings on that variable. Contrary to this assumption, confirmatory factor analyses revealed that the gold-standard items did not consistently exhibit the highest loadings on their corresponding latent variables, calling into question the content validity of our scales. Furthermore, when controlling for the self-reported physical body height, body weight, and age as gold-standard items, as well as gender, we did not always find an interpretable, systematic residual variance. Finally, the pattern of correlations among the latent variables did not closely reflect those among the self-reported gold-standard items, suggesting that the item scales and the gold-standard items do not share the same validity. While these results are consistent with previous studies, our analyses also showed that items with two response categories were as valid as those with six categories, contradicting past recommendations. Our analyses underscore the importance of intensive cognitive pretesting in scale construction and suggest that questionnaires using different response formats should be compared empirically and that hypotheses about which items best represent the latent variable should be tested.

In this study, we compared three different models of reading comprehension on a large dataset of more than 6,500 students. We compared two models representing four comprehension processes as well as the influence of text difficulty to a one-factor model by applying multi-dimensional item response models to the newly developed Bavarian reading test (BYLET). Cross-validation indicated the best generalizability for the one-factor model, but factor loadings and global model fit showed some evidence for the influence of text difficulty measured by word and sentence length and the process structure. All psychometric models tested had a good fit, as indicated by global fit indices and loading patterns. The general factor scores point to reliability and validity. We conclude that theories of reading comprehension processes also apply to some extent to the measurement of reading comprehension as a trainable skill and that the general factor score of the BYLET is suitable for a reading comprehension screening between grades two and six. The study is preregistered. The analysis code is available at https://osf.io/xw9bv/?view_only=21549993ef79426eb4006ef82415f25c . Test materials can be sent to interested researchers on request.

Hintergrund: Adaptiver Rechtschreibunterricht gilt wegen großer Unterschiede zwischen Grundschüler*innen als herausfordernd. Digitalisierung wird als "Gamechanger" diskutiert. Gleichzeitig erfordert eine Kultur der Digitalität die Erfahrungen und Herausforderungen der Schüler*innen mit multimodaler und -medialer Schriftsprache zu berücksichtigen. Beides hat das Potenzial, Unterricht zu transformieren. Ziel: In einem transdisziplinären Team wurde die App talidu entwickelt und untersucht, wie Erst- und Zweitklässler*innen in einer digitalen Lernumgebung Rechtschreibkompetenzen aufbauen sowie welche medien(un)spezifischen Strategien sie anwenden. Stichprobe: Aus einer Gesamtstichprobe von N = 685 Schüler*innen wurde nach Rechtschreibleistung das oberste und unterste Dezil ausgewählt, das mindestens 100 Wörter schrieb (ngesamt = 16). Methode: Unter Verwendung von Log-Daten wurden Extremgruppenvergleiche angestellt. Ergebnisse: Die Analysen zeigen geringe Unterschiede im Fehlerbild, aber deutliche bei den Korrekturvorgängen und der Nutzung von digitalen Hilfen. Diskussion und Fazit: Die Ergebnisse zeigen, dass Kinder beim digitalen Rechtschreiben konstruktiv aktiv sind, jedoch beim Testen von Rechtschreibstrategien, dem Korrekturverhalten und der Nutzung digitaler Hilfen unterstützt werden sollten. Die Ergebnisse verdeutlichen zudem den Informationsgehalt von Log-Daten und deren Bedeutung für den Rechtschreibunterricht im Allgemeinen. Dies unterstreicht das Potenzial digitaler Unterstützung z. B. durch Intelligente Tutorielle Systeme und zeigt Möglichkeiten, Unterricht in einer Kultur der Digitalität zu gestalten. (DIPF/Orig.)

We present TreaDeF, a formal, uncertainty-aware diagnostic treatment decision framework. TreaDeF is situated within statistical decision theory, merging traditional informal, human-centered and algorithmic decision making. TreaDeF extends existing approaches from psychometrics and econometrics by incorporating measurement uncertainty. It is a first approach towards individualized, transparent diagnostic decision making by incorporating personal characteristics and individual decision preferences. The framework extends the classical decision theoretic setup by estimating the utility function. While treatment options represent actions and personal characteristics, that determine treatment success, represent states; treatment success predictions for each combination of actions and states represent the utility function. Decision rules, like maximum expected utility or maximin applied to a probability-weighted utility criterion, complete the setup. Our implementation involves psychometric measurement models, non-linear prediction models, and numerically approximated decision rules. State uncertainty is determined by measurement error in Rasch models and utility uncertainty is determined by estimation error of generalized additive regression models. Finally, a numeric optimization method is used to apply maximum expected utility and maximin decision functions. A simulated example of a therapist deciding whether to administer exposition therapy is presented, as well as an extensive real-data example showcasing our implementation to a two-component educational training program for reading ability.

In expert–novice research paradigms on the teaching profession, the length of experience is—despite many warnings—used quite often as a key criterion for assigning teachers to the expert group. Using data from the COACTIV project (a study on the professional competence of mathematics teachers in PISA classes) we analyzed the relationship between 13 expertise indicators and length of experience (i.e., years of teaching). By ordering all COACTIV teachers regarding each of the 13 indicators and considering the median rank of a teacher according to these indicators, we were able to form an overall expertise (OE) index. When modeling linear correlations, the 13 indicators as well as the OE index showed negative or zero correlations with length of experience. When allowing for nonlinear models, we identified slight differential relationships between years of teaching and OE within three distinct intervals: a positive relation during the first 8 years of teaching, a negative relation in midcareer, and again a positive relation in the last 10–15 years. Similar curves also emerged in subsamples (academic- vs. vocational-track teachers). We discuss these findings in light of the cross-sectional nature of the COACTIV data. Our results call into question if long experience is a sufficient condition for becoming an expert teacher but strikingly also contradict the widely held view that professional experience is a necessary condition.

Parental socialization is the cradle of cultural learning, preparing young children for their specific lifeworld. Given rapid urbanization and cultural change worldwide, it is crucial to understand whether systematic differences in socialization practices exist. Urban as compared to rural contexts are theorized to differ in caregiving practices, yet systematic assessments across diverse contexts are still lacking. Here, we applied a novel behavioral socialization scale (BESOZ) to assess maternal behavior (N = 214) during interactions with their 3-year-old children across 13 standardized tasks, in three urban (Argentina, Germany, Japan) and three rural (Brazil, Mexico, Uganda) contexts worldwide. Latent profile analysis and machine learning revealed three patterns of maternal behavior and key behaviors that distinguished between contexts. Mothers in urban contexts provided suggestions, explanations, and positive feedback (profile 1), while mothers in rural contexts predominantly used directive prompts, demonstrations, and negative feedback (profile 2), or showed maternal dominance in task completion (profile 3). These findings reveal consistent differences in parental behavior across diverse urban and rural contexts, with urban mothers supporting children’s autonomy and independent thinking, and rural mothers capitalizing on social guidance and obligations. This contributes to a better understanding of early learning environments and a framework for studying parent-child interaction in a rapidly urbanizing world.

Objective: Teachers are at constant risk of straining their voices due to professional demands, making them more susceptible to voice disorders and therefore increased absenteeism. A widely accessible prevention method would be beneficial, leading to the development of the ReSt mobile application. The aim of this study was to evaluate its effectiveness. Methods: A randomized controlled trial was conducted with 399 teachers, allocating 205 participants to the training group and 194 to the control group. Over a 46-day period, participants in the training group engaged with the ReSt app. Outcomes were assessed using self-reports via the Voice Handicap Index (VHI-12) and objective evaluations employing the Dysphonia Severity Index (DSI). Additionally, app use was fully tracked using log data to gain insight into user engagement. Results: On average, the training resulted in substantial improvements in the DSI within the training group compared to the control group ( γ = 2.22 , S E = 0.32 , p < .001 ). No training effect was observed concerning the VHI-12 ( γ = − 0.39 , S E = 0.46 , p = .388 ). The average number of days spent opening exercises in the app was 10.57 days (SD = 9.07), and time spent on exercise-related app screens averaged 2.38 hours (SD = 2.79). Conclusions: The app appears to have promising effects on physiological measured voice capability, but more training time may be needed to minimize the psychosocial impact of voice problems. Our results mark an important advancement in voice disorder prevention approaches, especially given previously inconclusive evidence for traditional voice training methods.