Background: Random forest based variable importance measures have become popular tools for assessing the contributions of the predictor variables in a fitted random forest. In this article we reconsider a frequently used variable importance measure, the Conditional Permutation Importance (CPI). We argue and illustrate that the CPI corresponds to a more partial quantification of variable importance and suggest several improvements in its methodology and implementation that enhance its practical value. In addition, we introduce the threshold value in the CPI algorithm as a parameter that can make the CPI more partial or more marginal.Results: By means of extensive simulations, where the original version of the CPI is used as the reference, we examine the impact of the proposed methodological improvements. The simulation results show how the improved CPI methodology increases the interpretability and stability of the computations. In addition, the newly proposed implementation decreases the computation times drastically and is more widely applicable. The improved CPI algorithm is made freely available as an add-on package to the open-source software R.Conclusion: The proposed methodology and implementation of the CPI is computationally faster and leads to more stable results. It has a beneficial impact on practical research by making random forest analyses more interpretable.

In this article, the change in examinee effort during an assessment, which we will refer to as persistence, is modeled as an effect of item position. A multilevel extension is proposed to analyze hierarchically structured data and decompose the individual differences in persistence. Data from the 2009 Program of International Student Achievement (PISA) reading assessment from N = 467,819 students from 65 countries are analyzed with the proposed model, and the results are compared across countries. A decrease in examinee effort during the PISA reading assessment was found consistently across countries, with individual differences within and between schools. Both the decrease and the individual differences are more pronounced in lower performing countries. Within schools, persistence is slightly negatively correlated with reading ability; but at the school level, this correlation is positive in most countries. The results of our analyses indicate that it is important to model and control examinee effort in low-stakes assessments.

Aims: Cardiac rehabilitation plays a vital role in secondary prevention of cardiovascular patients. Female sex and higher age, however, are associated with non-referral to cardiac rehabilitation. Improving exercise capacity during cardiac rehabilitation is essential to reduce morbidity and mortality risks. The objective of this study was to closely examine the beneficial changes in exercise capacity of older patients of both sexes during cardiac rehabilitation and to identify the most important predictors of the change in exercise capacity. Method: A sample of 13,612 patients (mean age = 69.10 ± 11.8 years, 63.7% men, 19% > 80 years) was analysed. Data were prospectively assessed from 2012–2018 in six Swiss in-patient cardiovascular rehabilitation clinics. Improvement in exercise capacity measured with the six-minute walking test represents the outcome variable. Univariate and multivariate analyses, as well as the random forest method were used to estimate variable importance. Results: Mean improvement in the six-minute walking test was 113.5 ± 90.5 m (men = 118.7 ± 110.0; women = 104.4 ± 93.0, Cohen’s d = 0.16). The presence of heart failure, diabetes mellitus and psychiatric diagnoses was related to reduced but nonetheless clinically relevant six-minute walking test improvement. Random forest analysis suggests that baseline exercise capacity, age, time in rehabilitation and heart failure were the most important predictors for improvement in exercise capacity. Clinically relevant improvements in exercise capacity (>45 m) were also present into old age (85 years) and for both sexes. Conclusion: As indicated by these results, efforts need to be increased to refer eligible patients to structured rehabilitation programmes, irrespective of patients’ age and sex.

Children differ in their sensitivity to positive and negative environmental influences, which can be measured with the Highly Sensitive Child (HSC) scale. The present study introduces the HSC-21, an adaptation of the original 12 item scale with new items and factor structure that are meant to be more informative than the original ones. The psychometric properties of the HSC-21 were investigated in 1,088 children across Belgium and the Netherlands, including child and mother reports. Results showed evidence for (a) bifactor model with a general sensitivity factor and two specific factors (i.e., Ease of Excitation–Low Sensory Threshold and Aesthetic Sensitivity); (b) (partial) measurement invariance across gender, developmental stage, country, and informants; (c) moderate child–mother agreement; (d) good reliability; (e) normally distributed item scores; and (f) meaningful associations with personality and temperament across both samples. No evidence was found for HSC-21 as a moderator in the relationship between parenting and problem behaviors.