Slide 14 of 16
Notes:
The variance of the residuals defined by the parameters of ordinary regression lines of m1 vs m2 and m2 vs. m1 is minimised.
Limitations:
Cumbersome
The results are not easy to interpret
Provides a potentially new technique.
Kendall's Robust line fit
Compute all possible slopes using all i-th and j-th m1-m2 pairs. Median of these slopes is the "slope of the line" and the intercept is median of m1-b*m2
Limitations
Masked points are problematic
Nearly vertical points are not acceptable (infinite slope)
