I haven't been paying much attention to developments regarding the Wechsler Adult Intelligence Scale-
-Fourth Edition (WAIS IV), so today I took a peak. Information is available at the publishers web site
. Projected availability is summer 2008.
As expected, the revision will align the structure of the WAIS-IV to be consistent with the WPPSI-III and WISC-IV organization structures. Three new tests are available and samples can be viewed
at the site. I can't wait to see some CHC
-organized factor studies.
What having a look at WAIS (can't recall the version) today, as it happens, trying to work out exactly what the Comprehension subscale consists of. Do you happen to know of any good item analyses of performance? A quick eyeball suggests that all the items within Comprehension don't tap into the same thing.
The dimensionality of a set of items can be examined a number of ways. One needs to have the item level data. Using standard Rasch software one can extract a first principal component (assuming it is a feature of your software) to see if all items load on the same factor...and, if some don't, it will flag them as possibly measuring a different dimension. More sophisticated analysis can be done with MIRT (multidimensional itrem response theory) methods. These more sophisticated item analysis procedures tell you how many dimensions may be in a set of items and which items load on which dimensions.
Ooops - meant to say "Was having..." Anyway, yes, I'm aware of these sorts of statistical, bottom-up analyses. But I'm a fan of the sort of thing that, e.g., Earl Hunt (1974) and Just, Carpenter and Shell (199x, for some x?) did with Raven's matrices. Looking, spotting patterns, watching how people solve items, seeing if self-report reveals anything. Actually my favourite study on Raven's is by DeShon and colleagues (1995) where they use the verbal overshadowing effect to test their hypothesis that some items are mainly visual and some are mainly verbal-analytic. I wondered if you knew of any specific analyses of this form for WAIS? The nice thing about "IQ" tests is that there are loads of huge studies, so you could probably test dimensionality hypotheses quite nicely by looking at external correlations with other tests in a battery.
Nope. Not aware of any such analysis as you describe. Sorry.
Post a Comment