Page 180 - Second language development of newly arrived migrant kindergarteners - Frederike Groothoff
P. 180

180 Chapter 8 Measure of Lexical Richness (MLR) From the comparison between the consecutive models (see Table 8.6) it is apparent that a model for MLR with a fixed linear component – allowing for differences in Age – fit the data better than a model with only an intercept (Δχ2 (MLRb) = 19.25; df = 1; p < .001). Neither the variance within individuals (MLRc) nor the variance between individuals (MLRd) depends on the age of participants. Further, adding Age2 to the model did not improve the fit significantly (MLRe). Hence, in the final model (MLRb) a fixed effect of Age is needed, and with this model we continued the analysis22. Table 8.6: Fit of Different Polynomials (-2LL) for Changes in MLR score (114 cases) as well as the Comparison of Consecutive Models. Comparison -2LL Models ΔΧ2 Δdf 19.25 1 1.67 2 3.93 2 0.00 1 p <.001 .44ns .14ns .96ns      Model MLRa: β0ijcons a       537.59          MLRb: MLRa + β1Age1ij MLRc: MLRb + e1ijAge1ij MLRd: MLRc + u10jAge1ij MLRe: MLRd + β2Age2ij 518.33 MLRa vs MLRb 516.67 MLRb vs MLRc 512.74 MLRc vs MLRd 512.73 MLRd vs MLRe a In addition to the intercept, variance components for differences within and between individuals are estimated. Based on this General Development Model we constructed Figure 8.5, in which both the average development as well as the differences within and between individuals are represented (see Table 6.5 in Appendix 6 for the parameter estimates). The average MLR score at an age of 73 months was estimated as 2.73. Each month a child grew older, his MLR score increased by 0.09          22 The reliability of this model is low, .53, this might cause the non-significant improvement of the model when we allowed the variance between and within individuals depend on age, or the non-significant effect of the addition of Age2 in the model. 

   178   179   180   181   182