A critical assessment of sparse PCA (research): why (one should acknowledge that) weights are not loadings.

Principal component analysis (PCA) is an important tool for analyzing large collections of variables. It functions both as a pre-processing tool to summarize many variables into components and as a method to reveal structure in data. Different coefficients play a central role in these two uses. One focuses on the weights when the goal is summarization, while one inspects the loadings if the goal is to reveal structure. It is well known that the solutions to the two approaches can be found by singular value decomposition; weights, loadings, and right singular vectors are mathematically equivalent. What is often overlooked, is that they are no longer equivalent in the setting of sparse PCA methods which induce zeros either in the weights or the loadings. The lack of awareness for this difference has led to questionable research practices in sparse PCA. First, in simulation studies data is generated mostly based only on structures with sparse singular vectors or sparse loadings, neglecting the structure with sparse weights. Second, reported results represent local optima as the iterative routines are often initiated with the right singular vectors. In this paper we critically re-assess sparse PCA methods by also including data generating schemes characterized by sparse weights and different initialization strategies. The results show that relying on commonly used data generating models can lead to over-optimistic conclusions. They also highlight the impact of choice between sparse weights versus sparse loadings methods and the initialization strategies. The practical consequences of this choice are illustrated with empirical datasets.

Logistic regression with sparse common and distinctive covariates.

Having large sets of predictor variables from multiple sources concerning the same individuals is becoming increasingly common in behavioral research. On top of the variable selection problem, predicting a categorical outcome using such data gives rise to an additional challenge of identifying the processes at play underneath the predictors. These processes are of particular interest in the setting of multi-source data because they can either be associated individually with a single data source or jointly with multiple sources. Although many methods have addressed the classification problem in high dimensionality, the additional challenge of distinguishing such underlying predictor processes from multi-source data has not received sufficient attention. To this end, we propose the method of Sparse Common and Distinctive Covariates Logistic Regression (SCD-Cov-logR). The method is a multi-source extension of principal covariates regression that combines with generalized linear modeling framework to allow classification of a categorical outcome. In a simulation study, SCD-Cov-logR resulted in outperformance compared to related methods commonly used in behavioral sciences. We also demonstrate the practical usage of the method under an empirical dataset.

Probing the role of perception in fear generalization.

Behavior in novel situations is guided by similarities to previous experiences, a phenomenon known as generalization. Despite the widespread influence of generalization on healthy and pathological behavior, insight into the underlying mechanisms is lacking. It remains unclear whether a failure to notice situational changes contributes to the generalization of learned behavior. We combined a fear conditioning and generalization procedure with a perceptual decision task in humans and found that a failure to perceive a novel stimulus as different from the initial fear-evoking stimulus was associated with increased conditioned responding. These findings demonstrate the potential of a perception-centered approach to better understand (pathological) behavior and its underlying mechanism and are a promising avenue for the development of refined generalization protocols.

Revealing interdyad differences in naturally occurring staff reactions to challenging behaviour of clients with severe or profound intellectual disabilities by means of Clusterwise Hierarchical Classes Analysis (HICLAS).

BACKGROUND: Investigating interdyad (i.e. couples of a client and their usual caregiver) differences in naturally occurring patterns of staff reactions to challenging behaviour (e.g. self-injurious, stereotyped and aggressive/destructive behaviour) of clients with severe or profound intellectual disabilities is important to optimise client-staff interactions. Most studies, however, fail to combine a naturalistic setup with a person-level analysis, in that they do not involve a careful inspection of the interdyad differences and similarities. METHOD: In this study, the recently proposed Clusterwise Hierarchical Classes Analysis (HICLAS) method is adopted and applied to data of in which video fragments (recorded in a naturalistic setting) of a client showing challenging behaviour and the staff reacting to it were analysed. In a Clusterwise HICLAS analysis, the staff-client dyads are grouped into a number of clusters and the prototypical behaviour-reaction patterns that are specific for each cluster (i.e. interdyad differences and similarities) are revealed. RESULTS: Clusterwise HICLAS discloses clear interdyad differences (and similarities) in the prototypical patterns of clients' challenging behaviour and the associated staff reactions, complementing and qualifying the results of earlier studies in which only general patterns were disclosed. CONCLUSIONS: The usefulness and clinical relevance of Clusterwise HICLAS is demonstrated. In particular, Clusterwise HICLAS may capture idiosyncratic aspects of staff-client interactions, which may stimulate direct support workers to adopt person-centred support practices that take the specific abilities of the client into account.

Clusterwise HICLAS: a generic modeling strategy to trace similarities and differences in multiblock binary data.

In many areas of the behavioral sciences, different groups of objects are measured on the same set of binary variables, resulting in coupled binary object × variable data blocks. Take, as an example, success/failure scores for different samples of testees, with each sample belonging to a different country, regarding a set of test items. When dealing with such data, a key challenge consists of uncovering the differences and similarities between the structural mechanisms that underlie the different blocks. To tackle this challenge for the case of a single data block, one may rely on HICLAS, in which the variables are reduced to a limited set of binary bundles that represent the underlying structural mechanisms, and the objects are given scores for these bundles. In the case of multiple binary data blocks, one may perform HICLAS on each data block separately. However, such an analysis strategy obscures the similarities and, in the case of many data blocks, also the differences between the blocks. To resolve this problem, we proposed the new Clusterwise HICLAS generic modeling strategy. In this strategy, the different data blocks are assumed to form a set of mutually exclusive clusters. For each cluster, different bundles are derived. As such, blocks belonging to the same cluster have the same bundles, whereas blocks of different clusters are modeled with different bundles. Furthermore, we evaluated the performance of Clusterwise HICLAS by means of an extensive simulation study and by applying the strategy to coupled binary data regarding emotion differentiation and regulation.

Prospective study on the reversibility of neurobehavioral effects in workers exposed to manganese dioxide.

In 1987, a cross-sectional study in a dry-alkaline battery plant in Belgium revealed subclinical neurobehavioral dysfunctions associated with inhalation exposure to manganese dioxide (MnO2) particulate. The overall geometric mean of the time-weighted average concentration of manganese (Mn) in "total" dust (MnT) amounted, at that time, to 1 mg Mn/m3 and the duration of exposure was 5.5 years on average. An 8-year longitudinal investigation was conducted in this cohort (n = 92) in order to find out whether early effects on eye-hand coordination (EHC), hand steadiness (HST), and simple visual reaction time (VRT) were reversible when the airborne manganese concentration at the workplace was abated. During the observation period from 1988 to 1995, MnT monitoring was implemented on a monthly basis producing more than 1300 personal air samples, EHC tests were given yearly to assess the precision of the hand-forearm movement (PN1), and HST and VRT tests were carried out yearly since 1991. By the end of the study, the cohort size had dropped to 34 subjects. The model of unbalanced repeated measurements with unstructured covariance matrix and a time-varying covariate (log MnT) was the most appropriate to analyze the data. Wald chi 2 statistic was used for testing time-trends. The reduction of MnT over time was significantly associated with an improvement of the PN1 values (total cohort: Wald chi 2 = 8.5, p = 0.004; beta log MnT = -6.098 +/- 2.096). Like in the total cohort, time-trends were also found in the three exposure subgroups which could be identified in the cohort (average MnT over 1987-1992 were about 400, 600, and 2000 micrograms Mn/m3 for the low, medium, and high exposure subgroups, respectively). Only in the low exposure subgroup the PN1 value normalized when MnT (provisional estimates) decreased from about 400 to 130 micrograms Mn/m3 by the end of the study. Solely the reduction in MnT explained these findings on PN1, while a "healthy-worker-effect" mechanism was unlikely to have operated. The prognosis for the medium and high exposure subgroups remains uncertain as the improvement of their EHC performance may have been affected by past MnO2 exposure to such an extent that the persistence of a partial loss of EHC ability is suggested. The time courses of the HST and VRT test results, however, indicated the absence of any improvement, suggesting irreversible impairment of hand stability (postural tremor) and simple visual reaction time. A separate examination in a group of 39 control subjects, re-tested 10 years after the first test in 1987, virtually precluded age as confounding factor in this prospective study. The findings of the longitudinal study are corroborated by the outcome of a separate follow-up study in a group of 24 ex-Mn employees, who showed in 1996 a significant improvement of eye-hand coordination after at least three years with no MnO2 exposure; as to HST and VRT, there was no significant change in the deficit of these two neurobehavioral markers.

Urinary excretion of homovanillic acid in workers exposed to manganese.

Homovanillic acid, an end product of dopamine catabolism, and manganese (Mn) were measured in the urine of 68 male workers exposed to Mn-containing dust in a dry alkaline battery plant or an Mn oxide and salt producing plant, and in 35 control male subjects. The geometric mean of the airborne concentration of inhalable (total) dust amounted to 0.95 and 1.37 mg/m3 in the dry alkaline battery plant and the Mn oxide and salt producing plant, respectively. In the latter, a higher prevalence of increased values of urinary homovanillic acid concentration was found. In the total population, there was a low but statistically significant positive correlation between the concentration of homovanillic acid and Mn in urine (r = 0.20, P = 0.04) but there was no significant correlation between the level of homovanillic acid in urine and Mn in airborne dust or duration of exposure. This observation might be compatible with the stimulation of dopamine turnover in the brain, which has been observed in the early phase of Mn intoxication in animals. However, the large variability in urinary homovanillic acid excretion in control subjects precludes the use of this biological indicator to detect early interference of Mn with the dopaminergic system.

Fertility of male workers exposed to cadmium, lead, or manganese.

The effect of exposure to cadmium, lead, or manganese on male reproductive function was examined in 1988-1989 in Belgian blue-collar workers. The workers were exposed to cadmium in two smelters (n = 83; geometric mean urinary cadmium level = 6.94 micrograms/g of creatinine; mean duration of exposure = 24 years), to lead in a battery factory (n = 74; mean blood lead level = 46.3 micrograms/dl; mean duration of exposure = 10.7 years), or to manganese (manganese dioxide) in a dry alkaline battery plant (n = 70; median atmospheric concentration of total manganese dust = 0.71 mg/m3; mean duration of exposure = 6.2 years). Fertility in these workers and in an unexposed population (n = 138) was assessed by examining the birth experiences of their wives through a logistic regression model. The probability of a live birth was not different between the unexposed workers and the cadmium- or manganese-exposed workers before or after the onset of exposure. While the fertility of the lead-exposed workers was somewhat greater than that of the unexposed before the onset of exposure, a significant decrease in fertility was observed during the period of exposure to the metal (odds ratio = 0.65, 95% confidence interval 0.43-0.98).

Assessment of the permissible exposure level to manganese in workers exposed to manganese dioxide dust.

The prevalence of neuropsychological and respiratory symptoms, lung ventilatory parameters, neurofunctional performances (visual reaction time, eye-hand coordination, hand steadiness, audioverbal short term memory), and several biological parameters (calcium, iron, luteinising hormone (LH), follicle stimulating hormone (FSH), and prolactin concentrations in serum, blood counts, manganese (Mn) concentration in blood and in urine) were examined in a group of workers (n = 92) exposed to MnO2 dust in a dry alkaline battery factory and a matched control group (n = 101). In the battery plant, the current exposure of the workers to airborne Mn was measured with personal samplers and amounted on average (geometric mean) to 215 and 948 micrograms Mn/m3 for respirable and total dust respectively. For each worker, the lifetime integrated exposure to respirable and total airborne Mn dust was also assessed. The geometric means of the Mn concentrations in blood (MnB) and in urine (MnU) were significantly higher in the Mn exposed group than in the control group (MnB 0.81 v 0.68 microgram/100 ml; MnU 0.84 v 0.09 microgram/g creatinine). On an individual basis, MnU and MnB were not related to various external exposure parameters (duration of exposure, current exposure, or lifetime integrated exposure to airborne Mn). On a group basis, a statistically significant association was found between MnU and current Mn concentrations in air. No appreciable difference between the exposed and the control workers was found with regard to the other biological measurements (calcium, LH, FSH, and prolactin in serum). Although the erythropoietic parameters and serum iron concentration were in the normal range for both groups, there was a statistically significant trend towards lower values in the Mn exposed workers. The prevalences of reported neuropsychological and respiratory symptoms, the lung function parameters, and the audioverbal short term memory scores did not differ between the control and exposed groups. The Mn workers, however, performed the other neurofunctional tests (visual reaction time, eye-hand coordination, hand steadiness) less satisfactorily than the control workers. For these tests, the prevalences of abnormal results were related to the lifetime integrated exposure to total and respirable Mn dust. On the basis of logistic regression analysis it may be inferred that an increased risk of peripheral tremor exists when the lifetime integrated exposure to Mn dust exceeds 3575 or 730 micrograms Mn/m3 x year for total and respirable dust respectively. The results clearly support a previous proposal by the authors to decrease the current time weighted average exposure to Mn dust.

Urinary excretion of mercury after occupational exposure to mercury vapour and influence of the chelating agent meso-2,3-dimercaptosuccinic acid (DMSA).

The spontaneous and chelator mediated excretion of mercury in urine was investigated in male subjects occupationally exposed to mercury vapour (alkaline battery and chloralkali plants) who did not exhibit any sign of kidney damage. The time course of the spontaneous elimination of mercury in urine was examined in seven workers (age 22-40) who had been removed from exposure to mercury vapour (average duration of exposure 4.4 years) because their urinary mercury concentrations repeatedly exceeded 100 micrograms/g creatinine. The post exposure observation period started 10 to 29 days after the date of removal and lasted about 300 days (slow HgU elimination phase). For each worker, the kinetics of the spontaneous HgU decline followed a first order process; the biological half life ranged from 69 to 109 days (mean 90 days). The increased urinary excretion of mercury after a single oral administration of 2 g meso-2,3-dimercaptosuccinic acid (DMSA) was investigated in 16 control workers (group A; age 23 to 49), in 11 workers removed from exposure for at least two years (group B; age 27 to 41), and in 16 workers currently exposed to mercury vapour (group C; age 21 to 58). In group C, the DMSA experiment was repeated twice (three weeks before and three weeks after a holiday) after measures had been taken to reduce the mercury emission. The urinary mercury excretion was significantly higher during the 24 hours after DMSA administration in all groups compared with that in the 24 hours before. The bulk (50-70%) of the DMSA stimulated mercury excretion appeared within the first eight hours. In each group, the amount of mercury (microgram Hg/24h) excreted after DMSA was significantly correlated with that before administration of DMSA. The groups whose exposure had ceased, however, exhibited much higher correlation for coefficients (r=0.97 for group B and 0.86 for group C after three weeks of holiday) than those currently exposed to mercury vapour (r-0.66 for group C before and 9.58 after reduction of exposure). The data suggest that after a few days of cessation of occupational exposure to mercury vapour the HgU before and after administration of DMSA mainly reflects the amount of mercury stored in the kidney, which represents a mercury pool with a slow turnover.