[R-sig-Geo] ggwr residual types question and a few others
edzer.pebesma at uni-muenster.de
Mon Jan 25 10:53:57 CET 2010
Seth, I believe that response residuals are the residuals obtaind by
(measurement - prediction) on the measurement scale ([0,1], for logit);
working residuals are the residuals obtained in the last step of the
weighted minimization, i.e. on the transformed measurements. Note that
predictions are not 0 or 1, but usually somewhere between 0 and 1.
Individual elements of the object returned can be obtained by addressing
them explicitly. Finding out how they are called, e.g. by
for the additional fields, you probably need to specify more input
arguments (e.g. hatmatrix, fit.points). Look into the help page of gwr,
and please provide us with a reproducable example!
Seth J Myers wrote:
> First post, thanks for this list. I've just spent the last few days learning how to use ggwr in spgwr. I fit a glm with family=binomial(link="logit"), and have discovered how to export my SpatialPointsDataFrame as a text file so I can work with it in other packages I'm more familiar with (not ready to tackle the full R learning curve yet). I have a few basic questions first, for a logistic regression I understand pearson and deviance residuals, but what would the residual types of response and working be? I specified response thinking it would be binary response(0 or 1) - probability of 1, but it returned values that indicate this isn't the case. Also, in the spgwr documentation it states that ggwr returns a SpatialPointsDataFrame of class "gwr" and that its "data" slot can (will?) have the following: fit.points, weights, GWR coefficient estimates, R-sqaured, and coefficient standard errors. So, far the output I get from exporting mymodel to a txt using write.table or !
> viewed on the screen by simply mymodel$SDF only includes the following: row ID, intercept value, coefficient estimates for predictors, dispersion (always 1 in my case), working_resids, and x and y coordinates. Is there some 'switch' internal to ggwr function or something else to specify to get such things as SE of coefficients and (more importantly to me) a residual that truly is reponse(0 or 1)-prediction(0 to 1)? Thanks. -Seth Myers
> [[alternative HTML version deleted]]
> R-sig-Geo mailing list
> R-sig-Geo at stat.math.ethz.ch
Institute for Geoinformatics (ifgi), University of Münster
Weseler Straße 253, 48151 Münster, Germany. Phone: +49 251
8333081, Fax: +49 251 8339763 http://ifgi.uni-muenster.de
http://www.52north.org/geostatistics e.pebesma at wwu.de
More information about the R-sig-Geo