Once the factor analysis has done its job of organizing items into groups, it is time to see how well the groups of items hold together. This is the job of reliability analysis. Although there are many different reliability statistics, the most commonly used is the Cronbach’s alpha. The Cronbach’s alpha (with a Greek symbol of α) uses the associations among a set of items to indicate how well the items, as a group, hold together. Conceptually, the idea is that all of the survey items that are supposed to measure a single underlying construct should be answered in a similar way by respondents. This similarity of responses indicates that the construct is being measured reliably* by all of the items. On the other hand, if a person gives very different answers to items that are supposed to be measuring the same underlying construct, it is difficult to argue that these items offer a reliable measure of the construct. In a sense, a Cronbach’s alpha (more commonly referred to as the alpha) indicates the average associations among a set of items. Generally speaking, the more items there are in a reliability analysis, the higher the Cronbach’s alpha will be. After all, if two items have a correlation of r = .50, that is some evidence that the two items may represent an underlying construct. But if 8 or 10 items are all correlated with r’s of .50 or greater, then we can have a lot of confidence that these items measure one underlying construct. Similarly, if there are only 3 items, and one of them is not strongly correlated with the other two, the overall average correlation will be quite weak. But if there are 8 items and only one does not correlate strongly with the others, the overall average correlation will not be greatly reduced. So the strength of the alpha depends both on the number of items and on the strength of the correlations among the items. The strongest a Cronbach’s alpha can be is 1.0. A common rule of thumb is that when a set of...
Please join StudyMode to read the full document