Item Analysis

Item analysis ought to tell you what items you ought to keep and what items you ought to remove, in order to increase the internal consistency of the test (raise the reliability). It has nothing to do with validity. A "good" item contributes to the i nternal consistency reliability of the measure - nothing more.

Key Points

For a good item analysis, you ought to have 5 to 10 times as many subjects as you have items on the test.
Statistics for scale: the mean is just the average score on the test.
Item means: the mean is the average score on an item on the test.

Item-total statistics

These measure the relationship of individual test items to the composite score.

The corrected item-total correlation

This is the correlation between and item and the rest of the scale, without that item considered part of the scale. Without this correction, the correlation would be spuriously inflated, since it would count twice in the calculation of the correlation.
If the correlation is low, it means the item isn't really measuring what the rest of the test is trying to measure.

Squared multiple correlation

This measures how much of the variability in the responses to this item can be predicted from the other items on the test. If an item doesn't predict much of the variability, then you ought to drop it.

Alpha if item deleted

Cronbach's alpha or the other alpha ought to go up if you delete a spurious item. Deleting the item means you gain internal consistency.

Back to Statistics course

Lorraine Sherry
http://www.cudenver.edu/~lsherry/item_analysis.html
Updated April 1, 1997