Creates a Bland-Altman analysis including trending agreement from a dataset of simultaneous measurements with two methods (alt and ref).
Arguments
- df
Data frame. Each row contains simultaneous measurements of the same value.
- ref_col
name of the column containing reference measurements.
- alt_col
name of the column containing alternative measurements.
- id_col
name of the column containing unique subject id's.
- logtrans
Log-transform measurements before fitting the difference model.
- unit
Measurement unit (e.g. "L/min"). Currently only used for plots.
Details
compare_methods() uses lme4::lmer() to fit two mixed models:
A difference model describing the (dis)agreement between methods (alt - ref). This model is used to calculate the Bland-Altman statistics including trending parameters, accounting for repeated measurements between subjects.
A distribution model descibing the distribution of the measured value, represented by the average ((alt + ref)/2). This model is used to descibe within-subject and between-subject variation in the measured variable (e.g. cardiac output).
Examples
data(CO)
compare_methods(CO, ref_col = rv, alt_col = ic, id_col = sub)
#> 60 paired measurements in 12 subjects
#>
#> est
#> === Distribution ===
#> Mean : 5.035
#> Between-subject variation (SD) : 1.210
#> Within-subject variation (SD) : 0.283
#> Total variation (SD) : 1.243
#>
#> === Method comparison ===
#> Bias (alt - ref) : -0.705
#> Between-subject variation (SD) : 0.934
#> Within-subject variation (SD) : 0.413
#> Total variation (SD) : 1.022
#> Intraclass correlation
#> └ Between/Total variance : 0.836
#> Limits of agreement (95%)
#> ├ Upper limit : 1.298
#> └ Lower limit : -2.707
#> Percentage error : 0.398
#>
#> --- Trending ---
#> Within-subject perc. error : 0.161
#> Change LoA [±] (95%) : 1.146
if (FALSE) { # \dontrun{
compare_methods(CO, ref_col = "rv", alt_col = "ic", id_col = "sub") # also works
} # }