Page 49 - Never Too Far Away? The Roles of Social Network Sites in Sojourners’ Adjustment
P. 49

                                Across the three time points, the rank order of the three types of interaction in terms of frequency was similar; with FtF interaction with the host-country network as the most frequent (M1 = 4.17, SD = 1.05; M2 = 4.32, SD = 0.89; M3 = 3.98, SD = 1.20), followed by Facebook interaction with the host- country (M1 = 3.97, SD = 1.05; M2 = 3.98, SD = 1.01; M3 = 3.97, SD = 0.95), and lastly, Facebook interaction with the home-country as the least frequent (M1 = 3.83, SD = 1.10; M2 = 3.89, SD = 1.03; M3 = 3.66, SD = 1.01). The means of perceived social support across the three time points were: M1 = 3.24 (SD = 0.51); M2 = 3.30 (SD = 0.49); M3 = 3.29 (SD = 0.48); and for depressive symptoms were: M1 = 1.94 (SD = 0.56); M2 = 1.91 (SD = 0.57); M3 = 1.86 (SD = 0.52).
Model testing
We tested our proposed model in Figure 1 using cross-lagged (for the long- term effects) and non-lagged (for the short-term effects) reciprocal causal paths analyses. The non-lagged reciprocal analysis assumes causal effects that occur within a short span of time (Finkel, 1995; Kline, 2016). Although panel designs do not prove causality as conclusively as experimental designs, they are useful means to estimate reciprocal effects and assess whether a set of results is consistent with a causal model (Finkel, 1995). We conducted structural equation modeling (SEM) with the observed variables for the three types of interaction and the composite means for perceived social support and depressive symptoms, using Stata 14 and the maximum likelihood for missing values (MLMV) estimation method. Perceived social support and depressive symptoms were represented by composite scales rather than latent constructs in order to reduce the number of parameters to be estimated and ensure adequate power, also given the high reliability of these multiple indicator scales. We applied z-score standardization to ascertain the comparability of the coefficients across the variables. We also applied the assumption of constancy of structural effects and estimated the models by placing equality constraints on the parameters since: (a) the panel waves were equally spaced (Finkel, 1995); and (b) the participants were in varying stages of their studies at T1. We assumed equilibrium in the model such that “the changes in the underlying reciprocal causations have already manifested their effects and that the system is already in a steady state” (Kline, 2016, p. 137). Starting with the cross- lagged model, we constrained the autoregressive paths from T1 to T2, and from T2 to T3 to be equal and the corresponding cross-lagged causal paths across the waves to be equal. We also constrained the residual variances between variables
SNS, Support, and Depression 47
 






























































































   47   48   49   50   51