Dear Achim, thank you very much for the suggestions, they work well. Agreed that an ordinal logistic regression seems a more powerful choice of analysis method, and I'm using that also.
Thanks for providing the vcd package, it's proving quite helpful. Regards, Paul Paul Prew | Statistician 651-795-5942 | fax 651-204-7504 Ecolab Research Center | Mail Stop ESC-F4412-A 655 Lone Oak Drive | Eagan, MN 55121-1560 -----Original Message----- From: Achim Zeileis [mailto:achim.zeil...@wu-wien.ac.at] Sent: Friday, May 22, 2009 3:06 PM To: Prew, Paul Cc: r-help@r-project.org Subject: Re: [R] vcd package --- change layout of plot On Thu, 21 May 2009, Prew, Paul wrote: > Hello, > > I'm trying to use the vcd package to analyze survey data. Expert judges > ranked possible features for product packaging. Seven features were > listed, and 19 judges split between 2 cities ranked them. > > The following code (1) works, but the side-by-side plots for Cities PX, > SF are shrunk too much. Stacking PX on top of SF would make for a > better plot. (I could switch the order of Feature and Rank dimensions, > and go with the default side-by-side, but would prefer not to). Several comments: - Adding keep_aspect_ratio = TRUE does probably change what you call "shrunk too much". By default is aspect ratio is kept fixed for 2-way displays. - I would switch the splitting order of Feature/Rank because you probably want the distribution of ranks for a given feature and not the other way round. In that case, you might find etting split_vertical = TRUE aesthetically more pleasing. But it's probably a matter of taste. - Setting gp = shading_max is not really the best choice: If you would want a conditional independence test based on the maximum of residuals you should use panel = cotab_coindep. See the JCGS paper in the references and the accompanying vignette("residual-shadings", package = "vcd") for more details. But note that the double maximum test is not the most powerful test against ordinal alternatives (as one might expect due to the ordered Rank variable). hth, Z > (1) > cotabplot(~ Rank + Feature| Cities, data = Pack.dat, gp = shading_max, > rot_labels = c(90, 0, 0, 0),just_labels = c("left", "left", > "left", "right"),set_varnames = c(Feature = "")) > > Reading the vcd help, I got lost trying to understand the > panel-generating parameters I should use. My best guess was below (2), > but gave an error message. Clearly, I don't know what the paneling is > asking for. This is where I would like some advice, if anyone is > familiar with vcd. > > (2) > #### Tried to change the layout of trellis plot from horizontal to > vertical > Pack.mos<-mosaic(~Feature + Rank, data = Pack.tab, gp = shading_max, > rot_labels = c(0, 0, 0, 0),just_labels = c("left", "left", "left", > "right"),set_varnames = c(Feature = "")) > ## attempt to create an object for panel argument in cotabplot function > > pushViewport(viewport(layout = grid.layout(ncol = 1))) > pushViewport(viewport(layout.pos.row = 1)) > ## tell vcd to change the default layout, and what to put in the top > plot > > cotabplot(~ Feature + Rank | Cities, data = Pack.dat, panel = Pack.mos, > Pack.dat[["PX"]], gp = shading_max, rot_labels = c(0, 0, 0, 0)) > ## create mosaic plot that's conditional on Cities; first plot Cities = > PX > ## panel argument is an attempt to modify an example in the vcd help > file > > popViewport() > ## create the graphic > > Error: Cannot pop the top-level viewport (grid and graphics output > mixed?) > > # no point in gong on to code the plot for layout.pos.row = 2 > >> str(Pack.tab) > Error in `[.structable`(x, i, args[[2]]) : subscript out of bounds >> class(Pack.tab) > [1] "structable" "ftable" >> dim(Pack.tab) > [1] 7 2 7 > > Cities PX SF > Rank Feature > 1 Flexible 2 0 > Integrate.Probes 1 2 > Large/heavy 1 0 > Lockout 0 1 > Recyclable 3 5 > Rigid 0 0 > Small/light 2 1 > 2 Flexible 1 6 > Integrate.Probes 2 0 > Large/heavy 1 1 > Lockout 1 0 > Recyclable 2 0 > Rigid 1 0 > Small/light 2 2 > 3 Flexible 1 1 > Integrate.Probes 3 0 > Large/heavy 1 1 > Lockout 2 1 > Recyclable 1 3 > Rigid 0 0 > Small/light 0 3 > 4 Flexible 3 0 > Integrate.Probes 0 2 > Large/heavy 0 0 > Lockout 2 2 > Recyclable 0 1 > Rigid 1 2 > Small/light 3 2 > 5 Flexible 1 1 > Integrate.Probes 1 1 > Large/heavy 0 3 > Lockout 0 2 > Recyclable 2 0 > Rigid 3 1 > Small/light 2 1 > 6 Flexible 0 1 > Integrate.Probes 1 3 > Large/heavy 3 2 > Lockout 3 1 > Recyclable 0 0 > Rigid 2 2 > Small/light 0 0 > 7 Flexible 1 0 > Integrate.Probes 1 1 > Large/heavy 3 2 > Lockout 1 2 > Recyclable 1 0 > Rigid 2 4 > Small/light 0 0 > > > >> sessionInfo() > R version 2.9.0 RC (2009-04-10 r48318) > i386-pc-mingw32 > locale: > LC_COLLATE=English_United States.1252;LC_CTYPE=English_United > States.1252;LC_MONETARY=English_United > States.1252;LC_NUMERIC=C;LC_TIME=English_United States.1252 > > attached base packages: > [1] grid tcltk stats graphics grDevices utils datasets > > [8] methods base > > other attached packages: > [1] relimp_1.0-1 vcd_1.2-4 colorspace_1.0-0 > MASS_7.2-46 > [5] RSiteSearch_0.1-5 brew_1.0-3 lme4_0.999375-30 > Matrix_0.999375-26 > [9] lattice_0.17-22 Rcmdr_1.4-9 car_1.2-13 > > loaded via a namespace (and not attached): > [1] tools_2.9.0 > > > If someone can give advice on stacking the two cities' plots, I would be > grateful. > > Thanks, Paul > CONFIDENTIALITY NOTICE: =\ \ This e-mail communication a...{{dropped:12}} > > ______________________________________________ > R-help@r-project.org 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. > > CONFIDENTIALITY NOTICE: =\ \ This e-mail communication a...{{dropped:12}} ______________________________________________ R-help@r-project.org 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.