If the reliability study yields two measurements, and if the raters are considered to be fixed (rather than representative of a pool of raters), then reliability can be estimated by computing the product moment correlation between the two measures. This is the usual correlation statistic built into most computer programs and calculators. When the ratings are quantitative, the correlation is known as the Pearson correlation, and when the ratings are binary it is known as the phi coefficient. Regardless of what they are called, they are comparable to the ICR(3,1) version of the intraclass correlation described above. For the Z variables the Pearson correlation is rP = 0.83 and for the X variables in Table 4 the phi coefficient is rP = 0.47. These are very close to the ICR(3,1) values of 0.81 and 0.47 obtained on the same data.
Was this article helpful?