TPC-Journal-V6-Issue1

The Professional Counselor /Volume 6, Issue 1 67 Table 1. Regression of Self-Injurious Behavior Variables B Expected B r Aggression* 1.220 3.388 .141* Depression* 2.381 10.819 .319* Late Adolescence (age) .234 .791 -.037 Male* .543 .581 -.090* Caucasian .056 1.057 .028 Note: *Significant at p < .001, Nagelkerke R 2 = .240 Table 2. Regression of Extremely Aggressive Behaviors Variables Expected B B Male* 2.86 1.05 Depression* 1.92 .653 Note: *Significant at p < .001 YRBS 2009 Analysis In Table 3, the regression for self-injurious behavior is presented. Given the base rates of the two coded options, 83% of the sample choose not to involve themselves in possible self-injurious aggressive behaviors (intentional fighting resulting in injury); therefore, the best predictive strategy is to assume that, for every case, the subject will choose not to participate in fighting behavior that would likely result in injury requiring medical attention. In essence, the odds of someone engaging in aggressive self-injury are approximately 20% (Exp B = .205). In testing the predictive model of age, gender, race, depression and likelihood to engage in individual self-injury, results indicate that the model was significantly predictive at Χ 2 = 984.4, p < .001. The Nagelkerke R 2 = .110 is an indication that this model would only account for 11% of the variance in predicting self-injurious aggressive behaviors (intentionally fighting to result in injury). After adding the predictive model, 83% of cases were correctly classified, as opposed to an 80% classification rate prior to the addition of variables to the predictive model. The Hosmer and Lemeshow test was not significant (χ 2 = 18.83, p > .001), indicating that the predicted probabilities match the observed probabilities. According to the predictive model, if the participant were female, she would be .326 as likely to engage in aggressive self-injurious behavior as compared to males. A Wald Test was used to examine the true value of the parameter based on the sample and all were found to be significant at < .001.