Classic calibration Standard regression assumes x values have no error Orthogonal distance regression (ODR) assumes both x and y have the same error variance.

Presentation on theme: "Classic calibration Standard regression assumes x values have no error Orthogonal distance regression (ODR) assumes both x and y have the same error variance."— Presentation transcript:

Classic calibration Standard regression assumes x values have no error Orthogonal distance regression (ODR) assumes both x and y have the same error variance

Calibrating buoys and satellites (and errors) at the same time Peter Challenor SOC

New method Assume that the true values are fixed but unknown Estimate the x’s as well as the regression parameters Indeterminacy so fix one regression to slope 1 and intercept 0

Mathematically Standard regression: all error on y Orthogonal Distance Regression : error equal on x and y New method. Estimate true value x

Results NDBCJASONTOPEXUKCMENVISAT Intercept (m) 0.0-0.216 (0.014) -0.088 (0.007) 0.616 (0.029) 0.125 (0.007) Slope 1.0001.035 (0.005) 1.008 (0.002) 0.802 (0.011) 0.980 (0.002) Residual std dev (m) 0.1460.1840.0880.3920.088

Download ppt "Classic calibration Standard regression assumes x values have no error Orthogonal distance regression (ODR) assumes both x and y have the same error variance."

Similar presentations