Construct validity
Construct validity (i.e., whether the test measures the intended constructs and subconstructs)
is demonstrated, in part, by data from exploratory and confirmatory analyses. Loadings from
the first unrotated factor structure from a Principal Component Analyses for a four factor
(alphabetics, fluency, vocabulary, and comprehension) solution and a five-factor (the four
areas of reading plus assessment) solution are all large. The initial eigenvalues for the first
factors were 2.84 and 3.68, respectively, for the four-factor solution and the five-factor
solution, reflecting 71 % and 73 % of the variance. Initial eigenvalues for all other factors
were below 1. These results provide support for interpretation of the composite score
(RIKC). These values are not unexpected given that all items were written to assess
knowledge of teaching reading skills to adults. However, there is some evidence to support
interpretation at the scale level. For example, when an independent factor structure was
created (using principal Component Analyses, Varimax Rotation, with Kaiser Normalization)
the structure provides limited support for independent interpretation of scale scores. In
fact, the data show that all the items load as expected. That is, they load highly on the factors
that define the scales (0.88 or 0.89 for the four factor solution and 0.67 to 0.88 on the fivefactor
solution) and low and trivially (below 0.29) on the other scales. Confirmatory factor
analyses may follow exploratory analyses in the initial stages of test development to better
define test structure. For ARIK-A, both four- and five-factor solutions are acceptable based
on chi-square indices (CMIN/DF), and root mean square error of approximation (RMSEA)
statistics, although the normed fit index (NFI) data show that the four-factor solution appears
to be slightly better (see Table 3). However, based on a follow-up analysis, the difference