Test Information Curve

From BEARWiki
Jump to: navigation, search

As described earlier, the usefulness of a particular proficiency estimate depends upon its standard error. Smaller standard errors of measurement imply more interpretable results than larger standard errors. Unlike classical test theory in which the standard error of measurement is assumed to be the same for all respondents, ConstructMap Lite computes different standard errors for different proficiency estimates.

A test information index can be computed from the standard error of measurement for a specific proficiency estimate to ascertain the sensitivity of the instrument relative to a respondent’s proficiency. A given instrument may provide more information about respondents at one range of proficiency than at another range; for example, a particular test may provide considerable information about average students, and less information about students at more elementary or more proficient levels. This is due to the standard errors being smaller near the mean of ability levels for that instrument.

The test information function is computed from

  1. Select Reports - Item Reports - Test Information Curve from the menu.
  2. Complete the Test Information Curve display options, illustrated in Figure 1. Select a title for the chart, select whether all cases should be included or only those with no missing data, and select the logit range to be displayed. The proficiency estimate type from the Proficiency Estimation Options dialog window will be used. Click on OK to continue.
    Figure1. Test Information Curve Display Options dialog window.
    Figure 2 is an example of the test information curve for Example 2. It uses the standard errors of measurement from the maximum likelihood estimates of proficiency (MLE). Note that this instrument is most sensitive at a relatively narrow range of proficiencies, from approximately -1.0 to 0.0 logits. This implies that the instrument is better at differentiating respondents with proficiencies in the -1.0 to 0.0 range than those with proficiencies above or below that range.
    Figure 2. Test Information Curve for Example 2, complete cases only.
  3. The graphic is automatically saved as a jpeg file in the folder specified in step 2, but you can use File – Save As PNG to save it in the Portable Network Graphics format, which may produce a higher quality image for print or web pages and usually produces a smaller file.
  4. Print the graphic using File – Print.
  5. Close the window by clicking on the OK button.