2. Examine the output for each reliability analysis. Compare the obtained alpha with the alpha estimated if each particular item was deleted. Would the alpha increase if an item were deleted from the scale? If the answer is no, retain all items. If the answer is yes, you may consider dropping the item with the lowest item-total correlation from the final version of the scale. First, however, ask yourself the following questions:

• Would dropping the item increase the alpha substantially?

• Is there a logical reason the item seems different from the other items loading on this factor?

If an item was dropped from a dimension, rerun the reliability analysis and repeat the process. Note that alpha is improved by dropping items with low item-total correlations.