Sometimes, questionnaires may contain several questions (i.e. items) that ask similar things. Therefore it can be more useful and informative to analyse groups of questions together, rather than analysing individual questions separately. To do this it’s possible to form a new variable, called a global scale, made by adding up the scores of the questions that are related.
However, before grouping individual items together, it is necessary to discover whether they do in fact measure the same construct. This is done by calculating a reliability statistic for the global scale called Cronbach’s alpha, which measures how closely related the items are as a group.
In order for the global scale to have good reliability, the Cronbach’s alpha should be fairly high (a commonly used threshold is 0.7). The closer Cronbach’s alpha is to 1, the more likely is that the items in the scale measure the same underlying concept.
If the global scale has a good reliability, it makes a lot of sense to add, or find the mean of, the scores across each of the items to come up with the global scale.