False parsimony and its detection with GLMs

A search for a good parsimonious model is often required in data analysis. However, unfortunately we may end up with a falsely parsimonious model. Misspecification of the variance structure causes a loss of efficiency in regression estimation and this can lead to large standard-error estimates, producing possibly false parsimony. With generalized linear models (GLMs) we can keep the link function fixed while changing the variance function, thus allowing us to recognize false parsimony caused by such increased standard errors. With data transformation, any change of transformation automatically changes the scale for additivity, making false parsimony hard to recognize.     

Statistical morphology

This paper first introduces a parametric model for the generation of stationary random correlated binary sequences. The parameters of the model include the probability that a pixel is a binary one pixel and the length of the structuring element which dilates the initially spatially uncorrelated sequence. The spatial statistics of such eroded, dilated, opened and closed correlated binary sequences are derived in terms of the spatial statistics of the input binary sequence. Understanding of such one-dimensional processing is a precondition for understanding what happens in the more interesting two- dimensional case.     

Problem behaviors,substance abuse and sexual abuse in psychiatrically hospitalized adolescents with bulimia nervosa

Twenty-eight adolescent bulimics were compared to 201 psychiatrically hospitalized non-eating disordered patients. Non-eating disordered patients were found to be reliably more aggressive, delinquent, and under-controlled than patients with bulimia nervosa. Sexual abuse was found to be less prevalent among bulimic adolescents than comparison adolescents. No significant between group differences were achieved on any measure of alcohol/drug abuse. The behavioral profiles of substance abusing bulimics were highly similar to those of non-eating disordered patients. Three months follow-up of bulimic patients found that they were less depressed, had less somatization, and were less overcontrolled; however, they showed no reliable improvements in eating attitudes/behaviours or drug/alcohol use.     

Gender,gambling and problem gambling

With data from a 1989 Iowa survey (N=1,011), adult male and female respondents are compared on their problem gambling, its correlates, as well as their gambling behavior. Gambling behavior means its scope, frequency, wagering and leisure time spent at gambling. Women's gambling behavior was lower than that of men, due to their having a narrower scope of gambling behavior, but the genders were not significantly different on frequency, wagering and time spent at gambling. Women and men did not differ significantly on problem gambling. Problem gambling is measured as loss of control over gambling, and consequences due to gambling as well as gambling behavior. Women and men did differ significantly, however, on several predictors of problem gambling. Women's estrangement from a conventional lifestyle and integration into a social world of gambling appeared to help explain their problem gambling. Alcohol consumption appeared to be a more important predictor for men than women. The genders shared the attitude that the odds can be beat as well as being big spenders as predictors of their problem gambling. The results are interpreted with practitioners' efforts to prevent and treat problem gambling in mind.This research was partially supported by the Iowa Department of Human Services and the Na tional Institute of Mental Health (1 RJ01 MH50369-01A1). The authors are grateful to anonymous re viewers and the editor of this journal for many helpful suggestions.