The RA then coded the remaining 15 h, while the researcher randomly selected 5 h of these data to use to perform reliability calculations. This exercise yielded good levels of rater-agreement, according to Landis and Koch 's (1977) scales (Tables 4 and 5).