AC1: Gwet's AC1 inter-rater statistic for 2 raters special case.

In DiagnosisMed: Diagnostic test accuracy evaluation for health professionals

Description

Compute inter-rater or intra-rater agreement. Kappa, a common agreement statisitcs, assumes that agreement is at random and it's index express the agreement beyond the one observed at random. Gwet's AC1 statistic assumes that are agreement between observers are not totally at random such as Kappa. There will be some cases easy to agree in the condition absence, and some cases easy to agree in the condition presence and some that will be difficult to agree. Gwet also discuss that AC1 is way less prompt to a paradox that occurs with Kappa with exteme prevalences of the condition, where low Kappa are observed even with high crude agreement. TIn simulations studies, Gwet estimated bias for this statistic, therefore, for apropriate interpretation of AC1, it is necessary to subtract AC1 from a critical value on table 6.16 on Gwet's book, and then use Fleis table as bencmarking. See http://agreestat.com/.

Usage

AC1(tab, conflev = 0.95, N = Inf, print = TRUE)

Arguments

tab	k x k table which represents table(rater1,rater2), must have equal number of rows and columns and have the same levels in the same order.
conflev	Confidence Level associated with the confidence interval (0.95 is the default value).
N	population size which will be stick in standard error correction, N = Inf is no correction.
print	Logical. Should results be printed on the console?

Author(s)

Marcel Quintana and Pedro Brasil. Special thanks to Mr. Gwet for reviewing the code.

References

Gwet. Computing inter-rater reliability and its variance in the presence of high agreement. The British Journal of Mathematical and Statistical Psychology. 61. 29-48. May 2008.

See Also

package irr for many agreement statistics. epiR::epi.kappa for Kappa with confidence intervals. Also take a look at epiDisplay::kap for multi-rater multi-reader Kappa.

Examples

table3 <- matrix(c(118, 2, 5, 0), nrow = 2, ncol = 2)

AC1(table3)

# The same thing as...
x <- AC1(table3, print = FALSE)
x

DiagnosisMed documentation built on May 2, 2019, 5:21 p.m.