We conducted two conceptually interconnected analyses consistent with the two goals of this
study. In our effort to map current instructional leadership profile in Thailand, we analysed the
dataset collected in 2008. We began by examining the structure of the dataset and generating
descriptive statistics for principal instructional leadership on the three dimensions of the PIMRS
framework. Missing data (less than 1% of the data) were imputed using the expectation-maximization (EM) algorithm on the total sample (n ¼ 1195). Next, we employed latent mean
analysis, a form of structural equation modelling (Aiken et al., 1994), in order to examine patterns
of instructional leadership across sub-groups of principals within the sample. We employed latent
mean analysis rather than multivariate analysis of variance owing to its superior ability to detect
measurement error (Aiken et al., 1994; Cole et al., 1993; Hancock, 1997).