Page 154 - Never Too Far Away? The Roles of Social Network Sites in Sojourners’ Adjustment

2011). Given that this dissertation targeted a special population defined based on a transient, temporary and finite circumstance (i.e., short-term migration), there was much difficulty in getting a large sample of respondents and retaining the respondents over time for the longitudinal study. It was also a challenge to find sojourners engaged in long-distance romantic relationships. Note though that based on a systematic review of Zhang and Goodson (2011), the sample sizes of the 64 articles they reviewed ranged from 21 to 631, and these were mostly cross- sectional studies. The studies in this dissertation can be considered having a good sample size relative to the previous studies on sojourners, and sufficient for the statistical analyses to answer the research questions. Moreover, the participants in the studies came from different countries, and were based in different countries as well. Although not representative, the samples provide a relatively higher external validity since the participants were not only coming from a certain national group, nor studying in the same university/working in the same company in a single country. As with the attrition of participants in the longitudinal study, panel attrition analyses showed that there were no serious selection problems due to the loss of participants in the succeeding measurement occasions. Also, the assumption of constancy in structural effects across the three time points in the panel design was maintained despite attrition. Nonetheless, the number of participants limited statistical power and prevented more complex model testing. Moreover, the sample size limited latent variable modeling in the longitudinal panel study. Such analysis could clarify true effects by isolating measurement-related variance. Also, larger sample sizes are needed to perform multilevel analytical procedures that differentiate aggregate or group level effects from within-person variations (e.g., random intercept cross-lagged panel model). Lastly, the sample size limited the inclusion of potential moderators (e.g., public vs. private communication via SNS, co-national vs. host-national host-country network). Thus, future studies should ensure larger sample sizes to have greater flexibility in performing complex model testing, as well as attain greater statistical power.
Future studies should also consider the time and number of assessments in the longitudinal study. As of yet, there are no established theoretical or empirical bases for the choice of the time lag between measurements. In this dissertation project, I based the time interval on a previous longitudinal study on international and domestic student sojourners’ adjustment (Hechanova, et al., 2002). However, it is possible that a 3-month interval between assessments is not sufficient or even too long to detect changes. Thus, future research should look into both more fine-
152 Chapter 6