LPA: Modeling the Complexity and Heterogeneity
First, LPAs were performed to determine the number of profiles needed to reconstruct the number and variety of Mexican-heritage youth’s ethnic identity and cultural/linguistic orientations. Separate LPAs were run at each wave because whether the overall “big picture” would remain the same or change (e.g., a new profile emerging at later waves) was not clear. If the profile solutions were found to vary across waves, such speciation and/or extinction of profiles would have to be incorporated into the analysis of the longitudinal trend (i.e., LTA).
This series of LPAs pointed to a four-profile model as the optimal representation of the current study’s data across all four waves. To illustrate, increasing the number of profiles consistently resulted in sizable reduction in both BIC and AIC up to the fourth profile (ΔBIC3→4-profile ranged from −47.22 to −142.11, and ΔAIC3→4-profile ranged from −70.05 to −168.36). In addition, all four-profile models demonstrated adequate levels of entropy (.84–.88). Modeling a fifth profile did not improve the fit, except for Wave 3 where the five-profile model appeared to fit better than the four-profile counterpart in terms of BIC and AIC (but the entropy difference was negligible). However, conceptual inspection revealed that the fifth profile marginally differed from the fourth one, adding little to the heuristic value of the model and making it difficult to interpret each profile. In contrast, all profiles extracted in the four-profile model were clearly interpretable as discrete patterns of ethnic identity and cultural/linguistic orientations. Hence, the decision was made to maintain the four-profile model across all four waves. provides a summary of thus identified four profiles’ parameter estimates and prevalence across time.
Parameter Estimates of the Four Latent Profiles and the Estimated Prevalence Across Waves
The first three profiles appeared indicative of various patterns of integration. The first profile, labeled as Strong Ethnic Identification (SEI), showed strong ethnic identity in terms of both Exploration and Affirmation/Belonging, coupled with a high likelihood of having a bicultural orientation and a relatively balanced use of both English and Spanish. The second profile exhibited moderate levels of Exploration and Affirmation/Belonging, a relatively high likelihood of having a bicultural orientation, and a tendency to use more English than Spanish. This profile was labeled as Moderate Ethnic Identification (MEI). The third latent profile had an interesting combination of weak Exploration with strong Affirmation/Belonging and a high likelihood of having a bicultural orientation. This profile was labeled as Strong Affirmation/Weak Exploration (SA/WE). Finally, the fourth profile had characteristics indicative of strong assimilation: weak ethnic identity, a low likelihood of having a bicultural orientation, and a tendency to use English more than Spanish. This profile was labeled as Monocultural-Assimilation (MA).
LTA: Scrutiny Into the Trend of Ethnic Identity and Cultural/Linguistic Orientations
Based on the four-profile model identified above, an LTA with the measurement invariance restriction (Lanza et al., 2003
) was performed using Mplus
4.2. All indicators were restricted to operate in the same way across all four waves so that the same profile would represent the same characteristics. Validity of this restriction was tested by comparing two models—one with the invariance restriction and the other without. The BIC suggested that the restricted LTA model (BICrestricted
= 33,917.63) fit the data better than the unrestricted model (BICunrestricted
LPA revealed complex, if systematic, patterns of the trend underlying Mexican-heritage youth’s ethnic identity and cultural/linguistic orientation profile development (see for a summary of the latent transition probabilities). First, youth with the SEI or MEI profile were most likely to retain it across waves, and if they were to change their profile, the most likely transition was between these two profiles (i.e., the adoption of the MEI profile by SEI youth or vice versa). As for those with the SA/WE profile at earlier waves, the overall trend was still the same; the most likely pattern was the maintenance of the SA/WE profile. If they adopt a new profile, the first choice would be SEI, followed by MEI. Finally, youth starting off with the MA profile were found to be the least stable. Except for the Wave 1 to 2 transition, MA youth were more likely to adopt the MEI profile in subsequent waves than SEI, which appeared to be the modal choice for the adolescents with the other profiles.
Latent Transition Probability Matrix Across Waves
To further scrutinize these profile endorsement stases and transitions, the frequency of unique longitudinal patterns was examined. The most striking finding that emerged from this inspection was the strong tendency to maintain one’s profile. Almost 40% maintained either the SEI or MEI profile throughout the four waves. Another 20% or so exhibited “semistasis”: maintenance of the same profile across three consecutive waves. In short, the major trend among these Mexican-heritage youth seemed to be the maintenance or initiation of the exploration of their ethnic background (see for a visual summary).
Structure of stasis/transition patterns of ethnic identity and cultural/linguistic orientation profile endorsement by Mexican-heritage youth
Linking the Ethnic Identity and Cultural/Linguistic Orientation Profile With Demographics
Finally, a series of multinominal logistic regression analyses were performed to gauge the demographic characteristics of these ethnic identity and cultural/linguistic orientation profiles, using the Wave 1 latent profile endorsement as the criterion. The MA profile was set as the baseline category of the regression model. The results revealed that several demographic factors were significantly associated with the adolescents’ profile endorsement (see ).
Multinominal Logistic Regression Table With the Ethnic Identity and Cultural/Linguistic Orientation Profile as the Criterion
Boys showed a weaker tendency than girls to endorse the SEI or SA/WE profile over MA. There were no significant gender differences in terms of MEI profile. Those with lower SES (indicated by lunch status) tended to endorse SEI. Interestingly, youth’s own birthplace was not significantly associated with his or her profile endorsement. On the other hand, parents’ birthplace predicted their children’s profile such that youth with U.S.-born father or mother tended towards the SA/WE profile. An ad hoc analysis on the father’s and mother’s birthplace interaction effect revealed that this tendency was present only for the youth both of whose parents had been born in the United States (B = 1.45, p < .01, OR = 4.27). Adolescents who had the experience of visiting family or friends outside the United States within the past 3 years were about 6.5 to 7 times more likely to have the SEI or MEI profile over MA. Intriguingly, visitation within the past year showed no such effects; in fact, it was associated with a reduced likelihood to endorse the SA/WE profile, indicating some delayed effects. The length of the time spent in the United States and the best friend’s ethnicity/nationality were not associated with the youth’s profile endorsement.