Using AMOS 19, we tested configural, metric, scalar and factor variance invariance for two
principal groups (that is, secondary principals versus others principals in primary and K–12
schools). The descriptive statistics revealed that the data were non-normally distributed (for example,
skewed towards the higher end of the rating response scale; see Appendix 1). Therefore, we
employed a bias-corrected method for latent mean analysis by using bootstrapping in order to
adjust the parameter estimates, standard errors and effect sizes. Bootstrapping involved resampling
and replacing the original dataset 1000 times.
To address the second research goal, we sought to understand if there had been changes in the
self-reported instructional leadership practice of principals following implementation of the reform
law in 1999. We compared the instructional leadership profiles of the 491 secondary school principals
in the 2008 dataset with the profiles of the secondary school principals generated in the three
doctoral studies conducted prior to passage of the 1999 education reform. This comparison was
possible because all of the four studies employed the PIMRS. We used descriptive statistics and
then computed effect sizes to make this comparison.