Shrout and Fleiss (1979) Case 2:
(MSR - MSE)/(MSR + (MSC - MSE)/n), for raters who are a RANDOM
SAMPLE and results meant to generalise to other raters. The
(MSC - MSE)/n term charges SYSTEMATIC RATER DIFFERENCES against
reliability, because a future rater brings their own bias.
ICC(3,k) drops it and is always at least as large – reporting
Case 3 for sampled raters is the standard way to overstate
reliability, so both are returned.
Usage
morie_psy_icc2k(y, subject, rater)
Arguments
- y
ratings.
- subject, rater
identifiers; a complete crossed design.
Value
list: value (ICC(2,k)), icc_single, icc3k, rater_penalty,
k, n, MSR, MSC, MSE, case, method.
References
Shrout and Fleiss (1979), Case 2 and Table 4;
McGraw and Wong (1996), Psychological Methods 1:30-46.
Examples
y <- c(9, 2, 5, 8, 6, 1, 3, 2, 8, 4, 6, 8)
morie_psy_icc2k(y, rep(1:3, each = 4), rep(1:4, 3))$value
#> [1] 0.6764706
