Method Comparison Study
In this (ongoing) project, I’ve been asked to analyze a set of data from a clinical veterinary laboratory.
The aim is to understand wether a new diagnostic method (or technology) can be used interchangeably with another one, considered the gold standard.
To do so, I’ve put the initial data in R and proceeded with the analyses using a specific R package called “MethComp”.
MethComp
The package morphs input data into an R-object, used by the same to do different type of plots such as:
- Identity line plot
- Bland-Altman plot
and also different type of regression models (Passing-Bablock or Deming regressions).
Initial Data
This table simulates the rows and the columns of the initial data that I’ve used for the analyses.
| id | methods | replicate | variable1 | … | variable x |
|---|---|---|---|---|---|
| 1 | A | 1 | 4.2 | … | 0.02 |
| 1 | A | 2 | 4.5 | … | 0.04 |
| 1 | B | 1 | 5.1 | … | 0.5 |
| 1 | B | 2 | 5 | … | 0.1 |
| 2 | … | … | … | … | … |
| … | … | … | … | … | … |
| 118 | B | 2 | … | … | … |
As we can see, there were two observations for each used method.
Identity and B-A plots
Thanks to the package, some initial (but sometime exhaustive) plot can be drawn:
The title represents one of the x variables analyzed by the method/technology. The upper plot is the Bland-Altman plot, while the lower one is the Identity line plot.
