rel (version 1.4.2)

kra: Krippendorff's alpha

Description

Calculates Krippendorff's alpha as an index of agreement for nominal, ordinal, interval, or ratio scale data.

Usage

kra(data = NULL, weight = c("nominal","ordinal","interval","ratio"), 
    conf.level = 0.95, R = 0)

Arguments

data

A matrix with n subjects and m observations (n*m matrix).

weight

Data scale of ratings specified as "nominal, "ordinal", "interval", or "ratio".

conf.level

Confidence level of the interval.

R

Number of bootstrap replicates used to estimate the confidence interval.

Value

method

Analysis name

raters

Number of raters

sample

Sample size

na

Percent of missing values

est

Point estimate

lb

Lower confidence boundary

ub

Upper confidence boundary

cont.table

contingency table

data

analyzed data

Details

Krippendorff's alpha is a measure of observed disagreement relative to disagreement expected by chance which has the advantages of being applicable to multiple raters, different scale metrics, and incomplete data sets (Krippendorff, 2004 p.221-243). Noteworthy is that alpha by definition is zero when the observed disagreement equals the expected disagreement (i.e., when variance is absent which is the case for perfect agreement). The confidence interval of alpha is derived via bootstrapping because its distribution is unknown.

References

Krippendorff, K. (2004). Content Analysis: An Introduction to Its Methodology. Thousand Oaks, CA: Sage.

Examples

# NOT RUN {
#Sample data: 4 raters, 12 subjects, and 5 ordinal scale response categories.
data <- cbind(rbind(1,2,3,3,2,1,4,1,2,NA,NA,NA),
              rbind(1,2,3,3,2,2,4,1,2,5,NA,3),
              rbind(NA,3,3,3,2,3,4,2,2,5,1,NA),
              rbind(1,2,3,3,2,4,4,1,2,5,1,NA))

#Krippendorff's alpha for ordinal scale data
kra(data = data, weight = "ordinal", conf.level = 0.95, R = 0)
# }