HCD, 3.0%
MCD, 15.2%
AD, 49.2%
AB, 32.5%
time period, DLC theories, and longitudinal research regarding delinquency and offending (D’Unger et al., 1998; Jennings & Reingle, 2012; Moffitt, 1993; 2006; Nagin & Tremblay, 2005a; Piquero, 2008). Also, consistent with developmental research regarding physical
aggression were the shapes of the latent class trajectories. All four classes peaked at age six and decreased as the study sample aged, with the higher classes decreasing more drastically (Nagin
& Tremblay, 2005a; Piquero et al. 2012a; Tremblay, 2003; 2010). Additionally, it should be noted that relatively low percentage of participants within the HCD class and largest class membership within the AD class are congruent with prior research and theoretically espoused percentages of class membership (Jennings & Reingle, 2012; Moffitt, 1993, 2006; Piquero, 2008). The only inconsistency was that of the percentage of study participants within the abstaining class. As discussed in detail below, this group appears to be comprised largely of females, which may explain the larger than typical number of participants within this class. As noted previously, this also aligns with prior research acknowledging that females are more likely to abstain from physical aggression and fall within analogous classes in latent trajectory research (Loeber & Stouthamer-Loeber, 1998; Odgers et al., 2008). It should also be noted that many previous empirical efforts employed male only samples (Fontaine et al., 2009; Xie et al., 2009).
Other potential explanations and risk related factors are analyzed in greater detail below.
Nonaggressive rule-breaking. Figure 2 depicts the results of the latent class growth curve analysis with nonaggressive rule-breaking as the outcome indicator across five waves of data. As noted above, the fit indices suggest that the most appropriate model estimated four classes. According to Figure 2 and as delineated in Tables 11 and 12, class 1 is comprised of 48.7% of the study sample and reports comparatively average scores of nonaggressive rule-breaking, starting with an estimated mean of 2.43 at age six. While the overall slope of the curve
is rather stable across all waves of data, participants within this class report a slight increase from age 12 to age 14. Similar to physical aggression, the estimated mean scores at all waves of data collection mirror the average scores for participants within the complete study sample (compare to Table 6). As a result of the nature of this latent trajectory, this class may be labeled as average escalators (AE).
Class 2, includes 4.0% of the study sample and reports comparatively high scores of nonaggressive rule-breaking. Members of this class indicate estimated means scores of 7.64 at age six. However, the mean scores for this class increase consistently across all waves of data. At age fourteen the estimated mean score for nonaggressive rule-breaking peaks at 10.79.
Therefore, class 2 may be referred to as high chronic escalators (HCE).
Class 3 is made up of 16.4% of the study sample and reports estimated mean scores that are approximately half the rate of those the HCE class at all data points. Similar to HCE, the lowest estimated mean score of nonaggressive rule-breaking for class 3 occurs at age six (3.87).
The slope of the curve for this class also consistently increases as participants age. Consequently, the present research labeled class 3 as moderate chronic escalators (MCE).
The final class within this model contains 30.9% of the study sample and consistently low mean scores of nonaggressive rule-breaking. At age six class members report estimated mean scores of nonaggressive rule-breaking at 0.75. The slope of this class’ curve is rather stable and remains below one across all waves of data. As a result, this class was labeled as abstainers (AB).
Similar to the findings for physical aggression, the number of latent classes, shapes of trajectory curves, and sample percentages within each class for the nonaggressive rule-breaking model were consistent with the study hypotheses and prior empirical research. The results
substantiate the notion that there is heterogeneity in latent class trajectories of nonaggressive rule-breaking during late childhood and early adolescence. Further, model fit was achieved at four latent classes, which aligns with literature on nonaggressive rule-breaking, DLC theories, and longitudinal research regarding delinquency and offending (Burt 2012; D’Unger et al., 1998;
Jennings & Reingle, 2012; Moffitt, 1993; 2006; Nagin & Tremblay, 2005a; Piquero, 2008).
Also, consistent with developmental research regarding nonaggressive rule-breaking was the shape of the latent class trajectories. All four classes reported the lowest scores of nonaggressive rule-breaking at age six and increased as the study sample aged, with the higher classes (HCE and MCE) increasing more drastically (Burt, 2012; Hopwood et al., 2009; Tremblay, 2010).
Additionally, it should be noted that relatively low percentage of participants within the HCE class and highest percentage membership within the AE class are congruent with prior research regarding class membership (Jennings & Reingle, 2012; Moffitt, 1993, 2006; Piquero, 2008).
Also similar to physical aggression, the only inconsistency was that of the percentage of study participants within the abstaining class. As discussed in detail below, this group also appears to be comprised of an overrepresentation of females, which may explain the larger than typical number of participants within this class (Fontaine et al., 2009; Piquero et al., 2005). As noted above, other risk related factors are analyzed in greater detail below.
Combined. Figure 3 illustrates the results of the latent class growth curve analysis for the combined model estimating individual growth curves for both physical aggression and
nonaggressive rule-breaking as the outcome indicators across five waves of data. Again, the combined model allows for the estimation of individual slopes and intercepts for separate concepts within the same model taking into consideration the manner in which covariates influence both outcome measures. As noted previously, the fit indices suggest that the most
appropriate model estimated four classes. It should be acknowledged that the data points are plotted using sample means as opposed to estimated means for this figure. Estimate mean scores employ multiple imputation to deal with missing data points, while sample means exclude the missing data.
Figure 2. Latent Class Growth Curves for Nonaggressive Rule-Breaking (Estimated Means).
0 1 2 3 4 5 6 7 8 9 10 11
6 8 10 12 14
Mean NARB Scores