First stage of the conventional SAD validation: computes Lin's concordance correlation coefficient and its components for each rater under each condition.
Arguments
- x
An object returned by
sad_data().
Value
A tibble with one row per rater and condition and columns n,
rho_c, r, Cb, v, u, intercept and slope.
Details
\(\rho_c = r \times C_b\), where \(r\) is Pearson's correlation
(precision) and \(C_b = 2 / (v + 1/v + u^2)\) is the bias correction
factor (accuracy). \(v = \sigma_{est}/\sigma_{act}\) is the scale shift
and \(u = (\mu_{est} - \mu_{act}) / \sqrt{\sigma_{est}\sigma_{act}}\) is
the location shift, negative when the rater underestimates. Note that
epiR::epi.ccc() reports the location shift with the opposite sign.
The intercept and slope of the regression of estimates on actual values are
also returned.
References
Lin, L. I. (1989). A concordance correlation coefficient to evaluate reproducibility. Biometrics 45:255-268.
Examples
x <- sad_data(SADRustSoybean, wide = c(UN = "unaided", SAD = "aided"), group = "sad")
head(sad_ccc(x))
#> # A tibble: 6 × 12
#> rater group method condition n rho_c r Cb v u intercept
#> <fct> <fct> <fct> <fct> <int> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
#> 1 1 Old UN UN_Old 50 0.901 0.942 0.956 1.05 0.297 7.22
#> 2 2 Old UN UN_Old 50 0.906 0.923 0.982 1.18 -0.0998 -5.85
#> 3 3 Old UN UN_Old 50 0.829 0.915 0.907 1.34 0.346 0.259
#> 4 4 Old UN UN_Old 50 0.923 0.927 0.996 1.09 0.0301 0.211
#> 5 5 Old UN UN_Old 50 0.771 0.916 0.841 1.36 0.531 4.35
#> 6 6 Old UN UN_Old 50 0.869 0.897 0.969 1.18 0.191 2.31
#> # ℹ 1 more variable: slope <dbl>