Winsteps estimates unidimensional Rasch models, but many of the benefits of multidimensional modeling can be obtained through a Winsteps analysis. The Winsteps results are often easier to understand an use than those of a multidimensional analysis.
Identifying dimensions among the items. This can be done by inspecting the content of the items or through Table 23.
1. Do a standard unidimensional Rasch analysis
2. Use Winsteps Table 23.1 and other information to identify each item's "dimension"
3. Enter a code for the item's dimension into column 1 of the item label
4. Do the standard Rasch analysis
5. Winsteps Table 31, Differential Person Functioning, will give a measure for each person on each "dimension", all in the same frame-of-reference.
Example 1: I always have a problem in unidimensionality. How can I solve this problem in this data and others?
Unexplned variance in 1st contrast = 4.6 so that the eigenvalue of first contrast is noticeably greater than 2.0
Reply:
(1) is it truly a problem? There are many reasons for dependency in the data. For instance, if your items are groups using ISGROUPS=, then items in the same group (and so with the same rating scale) are slightly more dependent with each other than with other items.
(2) Look at Table 23.1.
Look at the content (wording) of the items at the top of the plot, and compare them with the items at the bottom of the plot.
What is in the top items that contrasts with what is in the bottom item? Are they on different dimensions (for instance, geography and history) or is the difference only superficial (for instance, response format or position on page)?
(3) Please look at the disattenuated correlations of the person measures at the bottom of Table 23.1:
Approximate relationships between the KID measures
PCA ACT Pearson Disattenuated Pearson+Extr Disattenuated+Extr
Contrast Clusters Correlation Correlation Correlation Correlation
1 1 - 3 0.1404 0.2175 0.1951 0.2923 < far from 1.0
If the correlations are near 1.0, then the person measures for the different "dimensions" are statistically the same. No action is needed.
(4) If the items really are on different dimensions and the person measures have a low correlation, then split the items into two subsets .
Please also see:
Dimensionality: contrasts & variances
Dimensionality investigation - an example
Dimensionality: when is a test multidimensional?
Item principal components/contrasts in Table 23 Identifies structure and dimensionality in response residuals
Table 23.0 Variance components scree plot for items
Table 23.1, 23.11 Principal components plots of item loadings
Table 23.2, 23.12 Item Principal components analysis/contrast of residuals
Table 23.3, 23.13 Item contrast by persons
Table 23.4, 23.14 Item contrast loadings sorted by measure
Table 23.5, 23.15 Item contrast loadings sorted by entry number
Table 23.6, 23.16 Person measures for item clusters in contrast. Cluster Measure Plot for Table 23.6.
Table 23.99 Largest residual correlations for items
Youtube video explaining Table 23
Example 2: I want to do a multi-dimensional Rasch analysis with Winsteps.
Multi-dimensional Rasch analysis is implemented directly in ConQuest. In Winsteps, the data for all the domains could be in one control file. Item weighting would be assigned to each item for each domain in separate IWEIGHT= files, one for each domain. Each domain would be analyzed separately with its own IWEIGHT= file, and person measures produced. Finally, the set of person measures (one for each person for each domain) would be summarized into one measure for each person by using a summarizing analysis, e.g., Q-mode analysis.
