Krippendorff's alpha reliability

The coincidence matrix counts every ordered pair of values within a unit, weighted by 1/(m_u - 1), so units coded by different numbers of coders are handled without imputation. The difference function is what encodes the level of measurement.

Usage

Krpalp(data, level = "nominal")

Arguments

data: Coders-by-units matrix; NA marks a missing value.
level: One of nominal, ordinal, interval, ratio.

Value

List with estimate, alpha, D_o, D_e, n_pairable, n_units, method.

Details

Formula: alpha = 1 - D_o / D_e with D_o = sum o_ck delta^2 / n and D_e = sum n_c n_k delta^2 / (n (n - 1)).

References

Krippendorff (2004), Content Analysis: An Introduction to Its Methodology, 2nd ed., Sage, ch. 11.

Examples

V <- c(1, 2, 3, 4, 5, 6, 7, 8)
Krpalp(V)
#> $estimate
#> [1] -1.110223e-15
#> 
#> $alpha
#> [1] -1.110223e-15
#> 
#> $D_o
#> [1] 1
#> 
#> $D_e
#> [1] 1
#> 
#> $n_pairable
#> [1] 8
#> 
#> $n_units
#> [1] 1
#> 
#> $method
#> [1] "alpha = 1 - D_o/D_e on the coincidence matrix, Krippendorff (2004) ch. 11"
#> 
#> attr(,"class")
#> [1] "morie_rich_result" "list"