Generation of SEM path models. Analyses described here used an iterative
process to develop single level SEM models for TDS. Researchers developed
separate single-level models for the classroom, school, and district-levels.
The development of each single-level path model began by specifying
a hypothetical path model (e.g., hypothesized model for TDS based on
classroom-level factors) based on prior research findings and on correlation
analyses; it was then estimated using AMOS. Since analyses presented here
aimed to develop a multilevel path mode and to estimate effects at different
levels within a school system, SEM analyses used covariance matrices.
AMOS provides several fit indices that provide estimates for the extent to
which the hypothetical model deviates from the theoretical model. Researchers
used overall fit statistics, absolute fit indices, and incremental fit
indices to evaluate model fit. Overall fit indices used were the chi-square
statistic (χ²) and the chi-square degrees of freedom ratio (χ²/DF). Absolute
indices used were the goodness of fit index (GFI) and the adjusted goodness
of fit index (AGFI). Incremental fit indices used were the normed fit index
(NFI) and the comparative fit index (CFI). Table 4 presents the evaluation
criteria used.