Denis Larocque

Negative Social Experiences and Dropping Out of School

The purpose of this study was to test a prediction model of early withdrawal from school with a sample of 751 low SES boys with an emphasis on the role of peer-related variables. The model included early disruptiveness and academic performance as the main triggering factors of a chain of events that included social acceptance from peers, and association with deviant/dropout friends. The model also included socio-familial variables, as well as parental variables as covariates and as potential moderators. Results confirmed the power of early disruptiveness and early academic performance to predict dropout. However, the effect of these variables varied according to dropout age. Low social acceptance from classmates made no additional contributions in predicting the process of early school disengagement. However, affiliation with deviant/dropout friends did. No moderating effects for parenting variables were found. Socio-psychological processes involved in the trajectory leading early disruptive boys to drop out of school and the importance of targeting early disruptiveness and academic difficulties to prevent school dropout are stressed in the discussion.

An empirical comparison of ensemble methods based on classification trees

In this paper, we perform an empirical comparison of the classification error of several ensemble methods based on classification trees. This comparison is performed by using 14 data sets that are publicly available and that were used by Lim, Loh and Shih [Lim, T., Loh, W. and Shih, Y.-S., 2000, A comparison of prediction accuracy, complexity, and training time of thirty-three old and new classification algorithms. Machine Learning, 40, 203–228.]. The methods considered are a single tree, Bagging, Boosting (Arcing) and random forests (RF). They are compared from different perspectives. More precisely, we look at the effects of noise and of allowing linear combinations in the construction of the trees, the differences between some splitting criteria and, specifically for RF, the effect of the number of variables from which to choose the best split at each given node. Moreover, we compare our results with those obtained by Lim et al. [Lim, T., Loh, W. and Shih, Y.-S., 2000, A comparison of prediction accuracy, complexity, and training time of thirty-three old and new classification algorithms. Machine Learning, 40, 203–228.]. In this study, the best overall results are obtained with RF. In particular, RF are the most robust against noise. The effect of allowing linear combinations and the differences between splitting criteria are small on average, but can be substantial for some data sets.

Verbal and Physical Abuse Toward Mothers: The Role of Family Configuration, Environment, and Coping Strategies

This paper focuses on parental maltreatment, a rarely documented phenomenon. We prospectively examine factors that can increase the risk of abusive behavior toward mothers. The purpose of this study is to examine the relationship between a stressful family situation, parental divorce, and verbal and physical aggression toward mothers. We use a large Canadian sample of adolescents who have been followed since kindergarten. At age 15, we assessed verbal and physical abuse toward mothers, as reported by both mothers and their adolescents. Multinomial logit modeling revealed that parental divorce was associated with a greater risk of physical aggression directed toward mothers by adolescents. Family environment and parental coping strategies partially mediated that relationship. Mothers who divorced, and remained divorced, were at greater risk of being assaulted by their adolescent children. A positive family environment, reflecting a better parent–child relationship, partially diminished this risk. However, support-seeking behavior on the part of mothers increased the risk of abuse, concurrent with tyrannical strategies often mobilized by abusive children.

The early explanatory power of NDVI in crop yield modelling

The objective of this paper is to study, on a weekly basis, the explanatory power of one satellite‐based measurement, the Normalized Difference Vegetation Index (NDVI), for wheat yield modelling in 40 census agricultural regions (CAR) in the Canadian Prairies during the whole growing season using 16 years of NOAA AVHRR satellite data (between 1987 and 2002). We also explore the relative value of NDVI compared with a land‐based measurement, the Cumulative Moisture Index (CMI). By developing a series of weekly wheat yield models over the course of the growing season, we are able to determine the accuracy of different models. Our findings indicate that NDVI possesses explanatory power 4 weeks earlier in the season than CMI.

Differentiating weak ties and strong ties among external sources of influences for enterprise resource planning (ERP) adoption

Enterprise resource planning (ERP) systems represent a major IT adoption decision. ERP adoption decisions, in the chemicals and allied products sectors, were examined between 1994 and 2005. Networks of strong ties and weak ties partners are investigated. Results show that neighbouring companies linked with strong ties can have an influence on organisations making such adoption decision. Past decisions made by major trading partners have a significant influence on the decision to adopt an ERP system for a given organisation. This reflects the complex nature of the knowledge required for such adoption.

The Wilcoxon Signed-Rank Test for Cluster Correlated Data

In this paper, we adapt the Wilcoxon signed-rank test to the case of cluster correlated data. A simple modification of the estimator of the asymptotic variance is sufficient to obtain a valid asymptotic procedure. However, the resulting test is no longer distribution-free. We derive the asymptotic null distribution of the statistic. A simulation study is performed in order to investigate the finite sample performance of the test. The results show that the performance of the test is very good for all designs and distributions considered when compared to competitors based on signs and on the overall average. In fact, the test is as powerful as the one based on the overall average for normal data as soon as intracluster correlation is present. An example of application with a real data set is also given.

Mixed-effects random forest for clustered data

This paper presents an extension of the random forest (RF) method to the case of clustered data. The proposed ‘mixed-effects random forest’ (MERF) is implemented using a standard RF algorithm within the framework of the expectation–maximization algorithm. Simulation results show that the proposed MERF method provides substantial improvements over standard RF when the random effects are non-negligible. The use of the method is illustrated to predict the first-week box office revenues of movies.

The impact of junior kindergarten on behaviour in elementary school children

Using data from the first cycle of the National Longitudinal Survey of Children and Youth, this study examines the impact of junior kindergarten on children’s behavioural development, above and beyond regional differences and individual/household factors. It was hypothesised that earlier schooling would help children learn behavioural control skills for participating in group settings like the elementary school classroom. Research has established that children from disadvantaged environments face an increased risk of behavioural problems, usually by virtue of accompanying risk factors. As such, we hypothesised that junior kindergarten attendance would reduce the risk gap for behavioural problems between children from economically disadvantaged and advantaged families. The results revealed that on the whole, junior kindergarten did not seem to decrease problem behaviour. These results are above and beyond a number of controls (sex, age, region, SES, family functioning, family configuration, education, and family size). Although children from disadvantaged environments exhibited more behaviour problems, attending junior kindergarten did not reduce the risk gap for behavioural difficulties between children from lower SES and higher SES backgrounds.

MACROECONOMIC EFFECTS OF TERRORIST SHOCKS IN ISRAEL

This paper estimates a structural vector autoregression to assess the empirical effects of terrorism on output and prices in Israel. Long‐run restrictions are used to interpret the effects in terms of aggregate demand and supply curves. The responses indicate that the immediate effects of terrorism are similar to those associated with a negative demand shock. Such a leftward shift of the aggregate demand curve is consistent with the existence of adverse effects on most components of aggregate expenditure documented in previous empirical studies. The long‐term consequences of terrorism are similar to those related to a negative supply shock. Such a leftward shift of the long‐run aggregate supply curve agrees with adverse effects on the determinants of the potential output, such as contractions of physical capital highlighted in earlier work, as well as reductions of technological innovations and slowdowns of net immigrations, which have not been fully analyzed in the existing empirical literature.

A weighted multivariate signed-rank test for cluster-correlated data

A weighted multivariate signed-rank test is introduced for an analysis of multivariate clustered data. Observations in different clusters may then get different weights. The test provides a robust and efficient alternative to normal theory based methods. Asymptotic theory is developed to find the approximate p-value as well as to calculate the limiting Pitman efficiency of the test. A conditionally distribution-free version of the test is also discussed. The finite-sample behavior of different versions of the test statistic is explored by simulations and the new test is compared to the unweighted and weighted versions of Hotellings T^2 test and the multivariate spatial sign test introduced in [D. Larocque, J. Nevalainen, H. Oja, A weighted multivariate sign test for cluster-correlated data, Biometrika 94 (2007) 267-283]. Finally, a real data example is used to illustrate the theory.