Girden and Kabacoff Critiques, Article Critique Example

This study written by Goldstein et al. is looking at explaining why undergraduates engage in self-destructive behavior; that is, what factors have a correlation in predicting self-destructive behavior in college students. In order to explore this question, the authors form a regression model composed of total self-harm (outcome variable), and the following independent variables: maltreatment history, personality, past year substance abuse, past year binge drinking. Before conducting regression analysis, the authors went through a series of baseline tests to understand if there were statistically significant differences at baseline. The authors conducted a series of chi-square analyses, followed by Fisher’s exact tests to evaluate whether the differences are significant. After looking at the relationship between the outcome variable (self-harm) and the correlates; the authors then undertook hierarchical regression analysis to understand how depressive symptoms, childhood maltreatment, personality, and substance use impacted self-harm.

Looking at the Girden outline for regression analysis, the sample used by the authors, 320 first-year college students (65% female), is representative to that cohort alone, and any generalizations beyond that cohort are not supported by the data. Indeed, the authors continually state throughout the article, particularly in the discussion section, that results from the sample is indicative of “undergraduate students.” This seems to be an error: While undergraduates are part of the undergraduate class, it seems that freshman would be facing challenges and volatility in their lives that may not be generalizable to all undergraduate students.

The initial tests to test baseline differences were reliable and valid: that is, the authors used a chi-square and Fischer’s test to make sure that the groups did not differ at baseline.

The main area for analysis in this paper is the use of hierarchical regression analysis to test the hypotheses. In particular, the ordering of entrance variables into the model is critical in obtaining the correct result: In this study, the authors’ rationale for entering depressive symptoms first was the purported correlation between the dependent and independent variable. The authors followed to enter childhood maltreatment subscale scores followed by personality variables and substance use. While the authors did offer a modicum of reasoning, the reasoning does not hold up; that is, there is not robust theoretical reason for this order , and thus it should be questioned. The authors do not use a stepwise regression methodology- thus, this criterion is not applicable in this study. Finally, the authors do not make explicit statements regarding examination of outliers and multicollineariy in the study’s methodology. Although this is (somewhat) commonplace in the psychology literature, this does not meet the standard established by Girden.

Finally, the authors found several significant correlates through the use of hierarchical linear regression. In particular, the authors identified depressive symptoms, physical neglect, emotional abuse, openness, sensation seeking, and drug use (past year) as significant correlates. While this is interesting, it still calls into question why they entered the variables in the order they did.