[R-sig-eco] Orthogonal distance, data points to regression
Rafael Maia
queirozrafaelmv at yahoo.com.br
Mon Jan 26 05:42:25 CET 2009
Dear list members,
I am analyzing a data frame consisting of two variables. There is a
trade-off between these two variables, and I have estimated a
regression line which represents the "upper bound" of the resolution
of this trade-off. You can basically imagine a scatterplot with data
points roughly distributed as a triangle, and the hypothenuse of this
triangle being this upper bound limit.
I would like, thus, to calculate the orthogonal distance from each
data point to this regression line. I have two problems; first, the
upper bound regression line does not use all the data. I assumed I
could get the distances by using an offset and extracting the
residuals. When I plot the lm() with the offset, the predicted line is
roughly equivalent to the upper bound, but the residuals are evenly
distributed around zero (I'd expect them to be mostly negative). The
second problem is that these residuals, from what I understand,
wouldn't be orthogonal distances, but the difference between fitted Y
and Y.
Any ideas on how to extract this information?
Many thanks in advance!
Abraços,
Rafael Maia
---
"A little learning is a dangerous thing; drink deep, or taste not the
Pierian spring." (A. Pope)
Laboratório de Comportamento Animal
Departamento de Zoologia, Universidade de Brasilia
http://www.unb.br/ib/zoo/rhmacedo/
More information about the R-sig-ecology
mailing list