In contrast, simply collapsing item data into a single total score requires item analysis to be disjoint from the evaluation of growth.
In addition to more common tools in scale development, IRT was used to examine individual item characteristics, to exclude items that provided little information and to collapse item categories.
Optimizations to this strategy can be made by collapsing items only for some critical areas of the matrix where the presence of biclusters is scarcer.
Four items were removed from the 16-item pool based on Rasch model analysis (three items whose response options were not ordered and could not be collapsed; one item based on experts' discretion).
That is, these production errors were higher than those demonstrated by the Standard group in Experiment 1, and dropped to the level demonstrated by the Standard group in Experiment 2. Similarly, when we examined final test accuracy collapsed across item type, we found a significant interaction between Group and Experiment (F 2, 264) = 8.59, P < 0.001, ( {eta}_p^2=0.06 )).
On this item, response categories 1 (not at all true) and 2 (hardly true) were reversed, and thus, categories 1 and 2 were collapsed for item #2 in all subsequent validity analyses.
There was a great deal of overlap, so we collapsed similar items into a single category, and chose one tool to represent them.
Using the RTT means, collapsed across items, as the dependent measure, we ran a 3-Time (Time 1, Time 2, Time 3) × 2-Intervention Group (Intervention First, Intervention Second) mixed ANOVA, with Intervention group as the between-groups factor, and time as the within-groups factor.
For this reason, options 3 and 4 of item 8 were collapsed, and that item was rescored 0-1-2-3-3-4, while the ofiginal scoring structure of 0-1-2-3-4-5 was retained fothers others.
For the retrieval session, collapsing across item type (i.e. studied and unstudied items), the data related to mean accuracy were analysed.
