We use maximum likelihood (ML) estimation in
AMOS 18 for both the measurement model and full structural model.
20
Among the 244 responses, most measures
have a full response, with no more than four responses missing for any single measure. AMOS does not evaluate missing
data, but provides a theoretical approach to random missing
data that are ‘‘efficient and consistent, and asymptotically
unbiased’’ (Byrne, 2001, p. 292). Where suggested by AMOS
and justified theoretically, we included covariances between
error terms of the same construct