ABSTRACT
Whereas the previous chapter presents the simplest α as an indicator of the reliability of data, this chapter highlights a few of α’s more general properties, repeating its numerical range. Between 0 and 1 α has valid reliability interpretations. Negative αs occur when disagreements are systematic.
To offer a sense of α’s scale, this chapter defines α’s degree of freedom and uses a very simple example of the number of possible coincidence matrices to show α’s scale to be one of equal intervals.
To make sense that various α-measures can be understood, this chapter associates its values with the proportion of information in the reliability data. α=1 when all data are unquestionably informative of the phenomena under consideration and α=0 when all of the data’s variance consists of unrelated noise.
This chapter shows α to be independent of the number of replications by observers, coders, or judges, and independent of the number of units observed, judged, and coded, which makes α unaffected by missing values – as long they are pairable within units of analysis.
The chapter concludes with arguing for the importance of variance in any effort to analyze data, echoing Section 1.2.4, and to measure their reliability as well.
