Bandwidth selection for local linear regression smoothers

Summary. The paper presents a general strategy for selecting the bandwidth of nonparametric regression estimators and specializes it to local linear regression smoothers. The procedure requires the sample to be divided into a training sample and a testing sample. Using the training sample we first compute a family of regression smoothers indexed by their bandwidths. Next we select the bandwidth by minimizing the empirical quadratic prediction error on the testing sample. The resulting bandwidth satisfies a finite sample oracle inequality which holds for all bounded regression functions. This permits asymptotically optimal estimation for nearly any regression function. The practical performance of the method is illustrated by a simulation study which shows good finite sample behaviour of our method compared with other bandwidth selection procedures.     

Mood in Daily Contexts: Relationship With Risk in Early Adolescence

Disturbances in affect have been linked to problem behavior in adolescence and future psychopathology, but little is known about how such disturbances manifest themselves in everyday contexts. This study investigated daily mood in Dutch 7th graders, aged 11–14. Cluster analysis of problem measures distinguished high-risk (n=25) and low-risk (n=106) subgroups. Participants completed experience-sampling reports of mood, social context, and location nine times daily for 5 days. Multilevel regression analyses of four mood measures confirmed higher anxiety and depressed mood in the high-risk group. Moods varied by location and social context, with significant differences between groups in two specific social contexts. First, when with family, low-risk adolescents felt less depressed than when alone, whereas high-risk adolescents felt more depressed. Second, high-risk adolescents showed more pronounced anxiety in social situations outside the network of family and friends. These findings point to everyday social contexts in which young adolescents with emotional and behavioral problems appear to be particularly vulnerable.     

Prüfungshonorare in Deutschland – Determinanten und Implikationen

The audit fees available since 2005 as a consequence of the Bilanzrechtsreformgesetz (commercial code reform act) show an extremely high supplier concentration on the German audit market of publicly listed companies. This justifies the assumption of audit fees above perfect market equilibrium prices. However, the complete collection and multivariate regression analysis of all relevant audit fee and client data document a 24 % decrease in audit fees in 2005 and an average 15 % decrease in the two subsequent years due to auditor changes. This can be interpreted as an indicator for price competition. In addition, the audit fee data shows that the audit fee decreases are only partially compensated by subsequent increases. As a consequence, there is no evidence for lowballing at present. Further time series are necessary to show if the current trend is sustainable.     

Unterrichtsqualität an einer besonderen Schule: Videobasierte Evaluation eines Oberstufen-Internats für leistungsstarke und hoch motivierte Schülerinnen und Schüler

Based on theoretical constructs and empirical work on quality of instruction, this contribution reports on the evaluation of a upper secondary boarding school for highly motivated and high-achieving students. The evaluation of instructional processes is based on high-inference video ratings of aspects of instructional quality. Additionally, aspects of cognitively activating teaching that are supposed to tap the full capabilities of the students are assessed as a higher standard for upper secondary teaching. Interviews with the teaching staff show a high level of acceptance of this approach to instruction as a means to cater to the needs of the high-achieving. As expected, comparing the teaching assessments with a reference sample of TIMSS and TIMSS-R videos shows a good quality of teaching in the areas of classroom management and individual learning support. For the different aspects of cognitive activation there are some lessons at a high level, while for some aspects the variation within the school is rather large.     

Estimation and selection procedures in regression: An L1 approach

The authors consider the problem of estimating a regression function go involving several variables by the closest functional element of a prescribed class _G that is closest to it in the L₁ norm. They propose a new estimator ? based on independent observations and give explicit finite sample bounds for the L₁distance between ?g and go. They apply their estimation procedure to the problem of selecting the smoothing parameter in nonparametric regression.     

Wild Bootstrapping in Finite Populations with Auxiliary Information

Consider a finite population u , which can be viewed as a realization of a super-population model. A simple ratio model (linear regression, without intercept) with heteroscedastic errors is supposed to have generated u . A random sample is drawn without replacement from u . In this set-up a two-stage wild bootstrap resampling scheme as well as several other useful forms of bootstrapping in finite populations will be considered. Some asymptotic results for various bootstrap approximations for normalized and Studentized versions of the well-known ratio and regression estimator are given. Bootstrap based confidence interval s for the population total and for the regression parameter of the underlying ratio model are also discussed     

Ecological and Motivational Determinants of Activation: Studying Compared to Sports and Watching TV

How can we enhance activation? Studying shouldbe a challenging, yet rewarding activity forstudents who intend to graduate. The Flowtheory (Csikszentmihalyi, 1990, 1997) predictsthat differential levels of perceived challengeand skill (flow) are related to optimizedmental states and increased activation.However, the influence of concurrent mentalstates and specific environmental cues for thisstate of optimal experience is unknown. In thisstudy we explore the contextual and subjectivedeterminants of flow in relation to activationin studying, and compare this with sports andwatching TV or listening to the radio. Method: 43 undergraduate students at theUniversity of Maastricht were assessed with theExperience Sampling Method for one week(Delespaul, 1995). At random moments 10 timeseach day subjects evaluated the social context,activities, and mood states. Analyses were donewith multilevel random regression techniques.Results: We replicated the predictedflow-related patterns in activation andemotions. While overall activation wasincreased in high challenging moments(β = 0.51; 95% CI: 0.36, 0.19), thiseffect was less pronounced during study(β = ?0.16; 95% CI: ?0.25, ?0.07). Skillslevels did not affect activation(β = ?0.01; 95% CI: ?0.06, 0.05).Concurrent emotions were independently andadditionally related to activation(Δχ² ₍₄₎ = 117.12,p < 0.0001). Unexpectedly, activation increasedwith demotivation (β = ?0.12; 95% CI:0.16, 0.07). We found highly significant andadditional effects of context for all theactivities (study: χ² = 732.63;p < 0.0001; R² = 0.30; active leisure:χ² = 753.40; p < 0.0001; R² = 0.31;and passive leisure: χ² = 555.86;p < 0.0001; R² = 0.24). Conclusions. TheFlow theory is a valuable model leading topredictions of optimal experience as well asactivation. However, the dynamics of activityengagement are more complex and related toconcurrent emotions and context. In the Dutchstudent culture, escaping boredom or compulsoryduties seems to drive individuals more thanpursuing flow.     

The four-dimensional symptom questionnaire (4DSQ): measuring distress and other mental health problems in a working population

In non-clinical (working) populations it is important to differentiate between general distress, on the one hand, and psychiatric symptoms—depression, anxiety and somatization—on the other hand. The Four-Dimensional Symptom Questionnaire (4DSQ) is a new instrument that measures these four symptom dimensions (Terluin, ). This study aimed to investigate the psychometric properties of the 4DSQ in a working population. A postal stress and health survey was sent to all employees of a Dutch telecom company, 51% of whom responded (N=3852). The mailing included the 4DSQ, a set of questionnaires concerning job stress (e.g. psychological demands), coping style (e.g. problem-focused coping, avoidant coping), and indicators of strain (e.g. emotional exhaustion, fatigue). Cronbach's α for the four sub-scales of the 4DSQ ranged from .79 to .90. Factor analysis showed four factors corresponding to the four scales of the 4DSQ: distress, depression, anxiety, and somatization. The validity of the 4DSQ was assessed using (partial) correlations with job stress, coping, and strain. As expected, the distress scale showed the strongest correlations with the indicators of strain, as well as with job stress and coping. In conclusion, the 4DSQ is a reliable and valid instrument that can be used in a working population to distinguish between stress-related symptoms and psychiatric illness.