Re: [R] Trellis Conditional contourplot
Finally I got the unreasonable contour plotted. http://www.geocities.com/useebi/data/kdecontour.jpeg The codes to get the plot, cdplot2d(survived ~ age + sibsp, data = titanic3) are in http://www.geocities.com/useebi/data/cdplot2d.txt One step that I couldn't accomplish as in cdplot is to eliminate those undefined conditional densities outside the ranges of the two covariates. The plot dosen't make sense for areas with no x and y observation. Probably, this is why the goal did not sound like anything reasonable. Anyway, I guess the naked eyes are pretty good to estimate the conditional density from the color coded scatterplot if nothing else works! marco Deepayan Sarkar [EMAIL PROTECTED] wrote: On 10/25/06, Marco LO wrote: OK! # The R codes to generate the Trellis plot are from the R help archive by Karl Ove Hufthammer panel.estcontplot - function(x, y, ..., pretty=TRUE, region=FALSE) { cest.gr=kde2d(x, y) cest.gr.lat=con2tr(cest.gr) panel.contourplot(cest.gr.lat$x, cest.gr.lat$y, cest.gr.lat$z, subscripts = seq(nrow(cest.gr.lat)), pretty=pretty, region=region, contour=TRUE, ...) } # The titanic3 dataset is from http://biostat.mc.vanderbilt.edu/twiki/pub/Main/DataSets/titanic3.xls with missing value imputations according to the book library(Hmisc) set.seed(17) xtrans - transcan(~ I(age) + pclass + sex + sibsp + parch + survived, data = titanic3, n.impute=5, pl=F, trantab=T, imputed=T, transformed=TRUE) titanic3[ ,'age'] - xtrans$transformed[ ,'age'] titanic3[ ,'survived'] - as.factor(titanic[ ,'survived']) library(lattice) xyplot(age ~ sibsp | survived, data = titanic3, panel=panel.estcontplot) Anyway, I realized that Trellis could only be conditional on the categorical variable while what I really want is conditional density plot of survived on the two covariates. That goal does not sound like anything reasonable to me, but I don't know much about categorical data analysis, so I'll let someone else answer. However, based on your first post where you implied that you expected 'densities' to 'add up to 1', I _will_ suggest this: if the phrase counting measure does not mean anything to you, you should seriously consider consulting a local statistician to whom it does. -Deepayan - [[alternative HTML version deleted]] __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] Trellis Conditional contourplot
Finally I got the unreasonable contour plotted. http://www.geocities.com/useebi/data/kdecontour.jpeg The codes to get the plot, cdplot2d(survived ~ age + sibsp, data = titanic3) are in http://www.geocities.com/useebi/data/cdplot2d.txt One step that I couldn't accomplish as in cdplot is to eliminate those undefined conditional densities outside the ranges of the two covariates. The plot dosen't make sense for areas with no x and y observation. Probably, this is why the goal did not sound like anything reasonable. You can do something pretty similar with ggplot: qplot(age, sibsp, . ~ survived, data=titanic3, type=c(jitter,2density)) (except clearly the density calculation is very different) Hadley __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] Trellis Conditional contourplot
OK! # The R codes to generate the Trellis plot are from the R help archive by Karl Ove Hufthammer panel.estcontplot - function(x, y, ..., pretty=TRUE, region=FALSE) { cest.gr=kde2d(x, y) cest.gr.lat=con2tr(cest.gr) panel.contourplot(cest.gr.lat$x, cest.gr.lat$y, cest.gr.lat$z, subscripts = seq(nrow(cest.gr.lat)), pretty=pretty, region=region, contour=TRUE, ...) } # The titanic3 dataset is from http://biostat.mc.vanderbilt.edu/twiki/pub/Main/DataSets/titanic3.xls with missing value imputations according to the book library(Hmisc) set.seed(17) xtrans - transcan(~ I(age) + pclass + sex + sibsp + parch + survived, data = titanic3, n.impute=5, pl=F, trantab=T, imputed=T, transformed=TRUE) titanic3[ ,'age'] - xtrans$transformed[ ,'age'] titanic3[ ,'survived'] - as.factor(titanic[ ,'survived']) library(lattice) xyplot(age ~ sibsp | survived, data = titanic3, panel=panel.estcontplot) Anyway, I realized that Trellis could only be conditional on the categorical variable while what I really want is conditional density plot of survived on the two covariates. Thus I tried to twist cdplot with kde2d to substitute density. http://www.geocities.com/useebi/data/cdplot2d.txt However, I still have no success. Would anyone help, please! Thanks marco Deepayan Sarkar [EMAIL PROTECTED] wrote: On 10/24/06, Marco LO wrote: Hello All, I've problems to interpret the contourplot for a binary response dependent variable. For example, I ploted the survived probability for the titanic3 dataset, http://www.geocities.com/useebi/data/CondenContour.jpeg Trellis displayed two panels of contour, one for dead and one for alive. However, the contours between the two panels didn't conform to the rule of prob(dead) = 1 - prob(alive). How I should read the contours? What I did wrongly? Hard to say, since we have no idea what you actually did. As the footer of every r-help message says: PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. -Deepayan Is there an example that I could follow? Thanks! marco - [[alternative HTML version deleted]] __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] Trellis Conditional contourplot
On 10/25/06, Marco LO [EMAIL PROTECTED] wrote: OK! # The R codes to generate the Trellis plot are from the R help archive by Karl Ove Hufthammer panel.estcontplot - function(x, y, ..., pretty=TRUE, region=FALSE) { cest.gr=kde2d(x, y) cest.gr.lat=con2tr(cest.gr) panel.contourplot(cest.gr.lat$x, cest.gr.lat$y, cest.gr.lat$z, subscripts = seq(nrow(cest.gr.lat)), pretty=pretty, region=region, contour=TRUE, ...) } # The titanic3 dataset is from http://biostat.mc.vanderbilt.edu/twiki/pub/Main/DataSets/titanic3.xls with missing value imputations according to the book library(Hmisc) set.seed(17) xtrans - transcan(~ I(age) + pclass + sex + sibsp + parch + survived, data = titanic3, n.impute=5, pl=F, trantab=T, imputed=T, transformed=TRUE) titanic3[ ,'age'] - xtrans$transformed[ ,'age'] titanic3[ ,'survived'] - as.factor(titanic[ ,'survived']) library(lattice) xyplot(age ~ sibsp | survived, data = titanic3, panel=panel.estcontplot) Anyway, I realized that Trellis could only be conditional on the categorical variable while what I really want is conditional density plot of survived on the two covariates. That goal does not sound like anything reasonable to me, but I don't know much about categorical data analysis, so I'll let someone else answer. However, based on your first post where you implied that you expected 'densities' to 'add up to 1', I _will_ suggest this: if the phrase counting measure does not mean anything to you, you should seriously consider consulting a local statistician to whom it does. -Deepayan __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] Trellis Conditional contourplot
Hello All, I've problems to interpret the contourplot for a binary response dependent variable. For example, I ploted the survived probability for the titanic3 dataset, http://www.geocities.com/useebi/data/CondenContour.jpeg Trellis displayed two panels of contour, one for dead and one for alive. However, the contours between the two panels didn't conform to the rule of prob(dead) = 1 - prob(alive). How I should read the contours? What I did wrongly? Is there an example that I could follow? Thanks! marco - [[alternative HTML version deleted]] __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] Trellis Conditional contourplot
On 10/24/06, Marco LO [EMAIL PROTECTED] wrote: Hello All, I've problems to interpret the contourplot for a binary response dependent variable. For example, I ploted the survived probability for the titanic3 dataset, http://www.geocities.com/useebi/data/CondenContour.jpeg Trellis displayed two panels of contour, one for dead and one for alive. However, the contours between the two panels didn't conform to the rule of prob(dead) = 1 - prob(alive). How I should read the contours? What I did wrongly? Hard to say, since we have no idea what you actually did. As the footer of every r-help message says: PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. -Deepayan Is there an example that I could follow? Thanks! marco __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.