Hello List:
I'm writing in R some code to produce plots for residual analysis and
diagnostics in linear regressions. An example of the plots produced is given
for downloading at http://dl.dropbox.com/u/25445316/res_plots.png . Regarding
the example plot, I'd like to point out that: 1) Tendency lines based on
lowess estimations are drawn as green continuous curves, 2) "Extra"
variables or components (variables not in the model formula, but in the original
dataframe or otherwise specified by the user) are plotted and labeled with an
"extra:" preffix. 3) Moderate and extreme atipical levels of residuals
are drawn as dashed yellow or red horizontal lines. The dashed green lines
represent the 0 mean residuals are supposed to have, according to model
assumptions.
This plotting function is part of a set of tools I'm developing for
spanish-sepaking pre-grad students. For more information, please consult
https://sites.google.com/site/unamatematicaseltigre/trabajos-de-estadistica .
The questions I'd like to put forth to the list are:
1) This plot is based on one I saw with similar dashed lines for atipical value
limits, but now I can't recall in which package I saw this. Does anyone
know of any package that produces very similar residual plots? I need to know
this in order to give due credit.
2) What do the experts think of the green trend lines? Are they effective for
spotting residuals/predictor variables relationships or are they misleading?
3) In the future, I would like to plot a grey shaded region approximately
corresponding to the dispersion cloud, in order to appreciate homocedasticity or
the lack thereof. Perhaps this could be done by calculating "upper and
lower limit trend lines"? Any ideas on how to do this?
Thanks in advance,
José Loreto Romero Palma
[[alternative HTML version deleted]]
