In light of these findings, the relative performance of common survey designs are discussed and design lessons are drawn in order to enhance the accuracy of item-specific consumption reporting and, consequently, measures of total household food consumption.
For dichotomous items, if H is calculated on the transposed data matrix then we obtain a measure summarizing the accuracy of item ordering within a scale [ 27].
Finally, we have not performed item response theory analysis as the English version of the AS-20 has undergone such analysis; addition of such analyses will be helpful to remove inappropriate items and hence to improve the accuracy of the items.
This can be deconfounded with two between-subjects comparisons which revealed that items first output on an earlier test were significantly more accurate than those output first on a later test (t [175.02] = 2.08, p = .04); however, the accuracy of reminisced items did not depend on when an item was first output (t [205.93] = 0.23, p = .82).82
Some of these comparisons can be made on a between-subjects basis, which allows us to control for the time of first output when comparing the accuracy of reminisced items from Group A with consistently recalled and forgotten items from Group B. In the between-subjects comparison, consistently recalled items again revealed higher accuracy than reminisced items (t [38.09] = 3.83, p < .001).001
The investigators recorded whether these items were present, and, when present, assessed the accuracy of these items compared to the patient chart as a reference standard.
In addition to the presence of absence of key items, our study also assessed, when present, the accuracy of these items compared to information contained in the hospital chart, giving a more complete assessment of the quality of the discharge summary.
Through the feedback provided by the focus groups, it was possible to substantially improve the accuracy of the items.
