So much data....so little time!
In a prior shameless plug, I briefly summarized the results of a recently published CHC-based confirmatory factor analysis study of a WJ-III/WISC-III cross-battery data set (Phelps, McGrew, Knopik & Ford, 2005). Following a favorite quantoid mantra ("there is more than one way to explore a data set"), I couldn't resist but conduct a more loosey-goosey (sp?) exploratory analysis of the data.
One of my favorite exploratory tools, given the Gv presentation of the multivariate structure of the data, is hierarchical cluster analysis (sometimes referred to as the "poor man's" factor analysis). Without going into detail, I subjected the data set previously described to Ward's clustering algorithm. As a word of caution, it is important to note that cluster analysis will provide neat looking cluster dendograms for random data....so one must be careful not to over-interpret the results. Yet, I find the looser constraints of cluster analysis and, in particular, the continued collapsing of clusters of tests (and lower-order clusters) into ever increasing broad higher-order clusters very thought provoking---the results often suggest different broad (stratum II) or intermediate level strata (as per Carroll's 3-stratum model).
I present the current results "as is" (click here to view or download). Blogsters will need to consult prior posts to glean the necessary pieces of information to interpret the CHC factor codes and names, the abilities measured by the WJ III tests, etc.
To say the least, some interesting hypothesis are suggested. In particular, I continue to be intrigued by the possibility of a higher-order dual cognitive processing model structure (within the CHC taxonomy) --that is, a distinction between automatic vs controlled/deliberate processing
Intelligent Insights on Intelligence Theories and Tests
