Do complex models increase prediction of complex behaviours? Predicting driving ability in people with brain disorders

Prediction of complex behavioural tasks via relatively simple modelling techniques, such as logistic regression and discriminant analysis, often has limited success. We hypothesized that to more accurately model complex behaviour, more complex models, such as kernel-based methods, would be needed. To test this hypothesis, we assessed the value of six modelling approaches for predicting driving ability based on performance on computerized sensory-motor and cognitive tests (SMCTests?) in 501 people with brain disorders. The models included three models previously used to predict driving ability (discriminant analysis, DA; binary logistic regression, BLR; and nonlinear causal resource analysis, NCRA) and three kernel methods (support vector machine, SVM; product kernel density, PK; and kernel product density, KP). At the classification level, two kernel methods were substantially more accurate at classifying on-road pass or fail (SVM 99.6%, PK 99.8%) than the other models (DA 76%, BLR 78%, NCRA 74%, KP 81%). However, accuracy decreased substantially for all of the kernel models when cross-validation techniques were used to estimate prediction of on-road pass or fail in an independent referral group (SVM 73-76%, PK 72-73%, KP 71-72%) but decreased only slightly for DA (74-75%) and BLR (75-76%). Cross-validation of NCRA was not possible. In conclusion, while kernel-based models are successful at modelling complex data at a classification level, this is likely to be due to overfitting of the data, which does not lead to an improvement in accuracy in independent data over and above the accuracy of other less complex modelling techniques.     

Larger IOR effects following forget than following remember instructions depend on exogenous attentional withdrawal and target localization

When words are onset in the visual periphery, inhibition of return (IOR) for a subsequent target is larger when those words receive an intervening forget instruction than when they receive a remember instruction Taylor (Quarterly Journal of Experimental Psychology, 58A, 613-629, 2005). The present study manipulated the allocation of endogenous and exogenous attention to assess the source of the forget > remember IOR difference. We determined that the forget > remember IOR difference likely arises from the differential withdrawal of exogenous-rather than endogenous-attention. Furthermore, this forget > remember IOR difference occurs only when a spatially compatible localization response is required; it does not occur when a simple detection response or a perceptual discrimination is required. This suggests that the forget > remember difference in the magnitude of IOR is not due to differences in perceptual/attentional processing. Instead, an instruction to remember or forget biases spatial responses in accordance with whether a location has previously contained relevant or irrelevant information. We suggest that directed forgetting in an item-method paradigm is not accomplished by changes in attention; rather, the changes in attention are coincident with changes in memory and may serve to bias later responses away from a source of unreliable information.     

Visual metrics: discriminative power through flexibility

An important stage in visual processing is the quantification of optical attributes of the outside world. We argue that the metrics used for this quantification are flexible, and that this flexibility is exploited to optimise the discriminative power of the metrics. We derive mathematical expressions for such optimal metrics and show that they exhibit properties resembling well-known visual phenomena. To conclude, we discuss some of the implications of flexible metrics for visual identification.     

The perceived structural shape of thin (wire-like) objects is different from that of silhouettes

Observers are faster at judging the position of convex vertices compared to concave vertices. This is believed to be due to an explicit representation of position for visual parts. The best evidence comes from comparing the same contours perceived as either figures or holes, because this is a pure figure ground reversal (Bertamini and Croucher, 2003 Cognition 87 33 - 54; Bertamini and Mosca, 2004 Perception 33 35-48). Specifically, an interaction is present between type of object (object or hole) and shape. One assumption is that the contour of a silhouette is perceived as the rim of a solid object. It follows that a different pattern should be found for thin (wire-like) objects compared to silhouettes. We confirm this difference in three experiments. We argue that this is due to the perceived parts when contours can be interpreted as self-occlusion rims.     

Cross-sectional and longitudinal relations among peer-reported trustworthiness, social relationships, and psychological adjustment in children and early adolescents from the United Kingdom and Canada

Trustworthiness was examined in children and early adolescents from two countries. In Study 1,505 children in the fifth and sixth school years in the United Kingdom (mean age = 9 years 7 months) were tested across an 8-month period. In Study 2,350 sixth- through eighth-grade Canadian children and early adolescents (mean age = 12 years 11 months) were tested across a 1-year period. Participants completed measures assessing trustworthiness (ratings of promise and secret keeping by peer group and also by best friend in Study 1), assertiveness (Study 1), prosocial behavior (Study 2), social relationships (friendships and peer preference), and psychological adjustment (self-esteem, depressive symptoms, and loneliness). Peer-reported trustworthiness was positively associated with, and predicted changes across time in, the number of friendships. The observed relations were found to be statistically independent of peer preference and peer-reported assertiveness or prosocial behavior. Trustworthiness was found to be positively associated with peer preference (Study 1 and Study 2) and psychological adjustment (Study 1).     

Eyeblink conditioning deficits indicate timing and cerebellar abnormalities in schizophrenia

Accumulating evidence indicates that individuals with schizophrenia manifest abnormalities in structures (cerebellum and basal ganglia) and neurotransmitter systems (dopamine) linked to internal-timing processes. A single-cue tone delay eyeblink conditioning paradigm comprised of 100 learning and 50 extinction trials was used to examine cerebellar timing circuits in 13 medicated patients with schizophrenia and 13 age- and sex-matched controls. Patients with schizophrenia showed impaired learning of the conditioned response compared to controls and also greater within-subject variability in the timing of their responses. These findings are consistent with models of schizophrenia in which timing deficits underlie information-processing abnormalities and clinical features of the disorder.     

Neural correlates of lexicon and grammar: evidence from the production, reading, and judgment of inflection in aphasia

Are the linguistic forms that are memorized in the mental lexicon and those that are specified by the rules of grammar subserved by distinct neurocognitive systems or by a single computational system with relatively broad anatomic distribution? On a dual-system view, the productive -ed-suffixation of English regular past tense forms (e.g., look-looked) depends upon the mental grammar, whereas irregular forms (e.g., dig-dug) are retrieved from lexical memory. On a single-mechanism view, the computation of both past tense types depends on associative memory. Neurological double dissociations between regulars and irregulars strengthen the dual-system view. The computation of real and novel, regular and irregular past tense forms was investigated in 20 aphasic subjects. Aphasics with non-fluent agrammatic speech and left frontal lesions were consistently more impaired at the production, reading, and judgment of regular than irregular past tenses. Aphasics with fluent speech and word-finding difficulties, and with left temporal/temporo-parietal lesions, showed the opposite pattern. These patterns held even when measures of frequency, phonological complexity, articulatory difficulty, and other factors were held constant. The data support the view that the memorized words of the mental lexicon are subserved by a brain system involving left temporal/temporo-parietal structures, whereas aspects of the mental grammar, in particular the computation of regular morphological forms, are subserved by a distinct system involving left frontal structures.     

Four studies on how past and current suicidality relate even when "everything but the kitchen sink" is covaried

T. E. Joiner's (2004, in press) theory of suicidal behavior suggests that past suicidal behavior plays an important role in future suicidality. However, the mechanism by which this risk is transferred and the causal implications have not been well studied. The current study provides evaluation of the nature and limits of this relationship across 4 populations, with varying degrees of suicidal behavior. Across settings, age groups, and impairment levels, the association between past suicidal behavior and current suicidal symptoms held, even when controlling for strong covariates like hopelessness and symptoms of various Axis I and II syndromes. Results provide additional support for the importance of past suicidality as a substantive risk factor for later suicidal behavior.     

Income inequality and risk of suicide in New York City neighborhoods: a multilevel case-control study

Evidence on the relationship between income inequality and suicide is inconsistent. Data from the New York City Office of the Chief Medical Examiner for all fatal injuries was collected to conduct a multilevel case-control study. In multilevel models, suicide decedents (n=374) were more likely than accident controls (n=453) to reside in neighborhoods with greater income inequality even after controlling for individual characteristics; this relation was modified by age with an effect overall and among decedents aged 15-34 but not among decedents 35-64. These data suggest that income inequality may contribute to the risk of suicide in younger adults.     

Aktuelle Probleme der Krankenhausplanung in der Psychosomatischen Medizin und Psychotherapie

Ohne Zusammenfassung

---