4.2 Information contained
The itembank should hold all relevant information that may be needed to select items for a test. The item name and label are the basic information in the itembank. In the context of an Item Response model, and Rasch model, that is used for the \(D\)-score, the most important item feature is item difficulty.
For the itembank for the \(D\)-score, that is used to constructs instruments to measure development, other important features relate to the criteria of better measurement as described in chapter @(ch:criteria)). The purpose of the instrument and corresponding instrument criteria differ per study design. For example to measure development in a population as part of a population survey, the milestones may need to be caregiver reported whereas for individual monitoring, direct observation might be preferred.
The \(D\)-score itembank contains the following information:
Item The item name is the short unique name for the specific item. The naming scheme we are using for the GSED project holds information about other item features. The name has 9 characters, the first three are the original instruments name (e.g. by1 for Bayley 1; aqi for the ASQ), character 4 and 5 represent the domain the item was placed under in the original instrument, character 6 is a c for caregiver report or a d for direct assessment and character 7 to 9 is the item number in the original instrument.
tau The item difficulty (tau) is the calibrated item parameter that results from a \(D\)-score model. This parameter is used to calculate the \(D\)-score when the item is used in an instrument. The item difficulty is also used for selecting items.
Key The key indicates a specific model calibration (i.e. Rasch model) and has a specific set of estimated item difficulties. Currently, three keys are included: the dutch calibration, the gcdg calibration and the gsed calibration.
| Key | Model | Quadrature | Instruments | Direct/Caregiver | Reference |
|---|---|---|---|---|---|
"dutch" |
75_0 |
-10:80 |
1 | direct | (van Buuren 2014) |
"gcdg" |
565_18 |
-10:100 |
14 | direct | (Weber et al. 2019) |
"gsed" |
807_17 |
-10:100 |
20 | mixed | GSED Team, 2019 |
For new data, the advice is to use the most general key, "gsed".
Label The item label holds the actual content of the item. Mostly this corresponds to how the item finally is asked in the test instrument. When different translations of the item are available, new columns may be added with translations of the label.
Domain The domain column indicates the developmental domain that the item loads to. In practice, items may load to multiple domains so the domain in the itembank may represent the most dominant domain for that item.
Mode Items that measure early child development can be either directly observed (e.g. ) or caregiver reported (e.g. ).
Instrument The instrument column indicates the developmental instrument the item originates from. The gsed calibration contains items from 20 original instruments.
Materials A materials column can hold information on the materials needed to administer the item. For example the item “stack two blocks” requires the availability of two blocks when the item is directly observed. Caregiver reported items generally do not require any additional materials.
Equate In the equate column a group name is indicated. Some items are grouped together in an equate group because they measure the same thing in a slightly different way (see booklet II: chapter 4 for more detail). The items in the equate group have the same difficulty, so When items are selected for an instrument, only one item from the equate group needs to be selected.