向 scatter3d 图添加图例

Question

One of the possible packages for interactive 3D plots is rgl. What I would like to do is to build 3D scatterplot with color coding according to some factor variable. The 3D dimensional scatterplot is for resulting loadings from plsr analysis.

The resulting plot looks like

The example data is given in a table:

> loadings

      |      Comp 1         |        Comp 2            |    Comp 3           | Class
-------------------------------------------------------------------------------------------                    
TEMP  | -0.0607044182964255 | "0.0437618450165671"     |"0.045124991801441"  | "global"  
MW    | "-0.13414890573833" |   "-0.0970537799069731"  |0.263043734662182"   | "local" 
DM    |"-0.183751529577861" |  "-0.102703237685933"    |"0.0640549385564205" | "global" 
CHG   |"-0.0558781715833019"| "0.125155347350922"      |"-0.119258450107321" | "local"

or can be generated:

loadings <- data.frame(Comp1 = c(1.2, 3.4, 5.6, 13.1), Comp2 = c(4.3, 1.2, 7.7, 9.8),
                       Comp3 = c(1.2,6.9,15.6,15.0), 
                       row.names = c("TEMP", "MW", "DM", "CHG"), 
                       Class = c("global", "local", "global", "local"))
scatter3d(x=loadings[[1]], y=loadings[[2]], z=loadings[[3]], 
          point.col = as.numeric(as.factor(loadings[,4])), size = 10)

Obtained plot has the same style, but is much simpler as there are only two levels of variable "Class": 'global' and 'local'

Question is : Is it any possibility to add legend within rgl or maybe some independent legend can be attached to the plot? Thanks in advance for help!

Answer is:

scatter3d(x=loadings[[1]], y=loadings[[2]], z=loadings[[3]], 
          point.col = as.numeric(as.factor(loadings[,4])), size = 10, type = 's')
text3d(x=1.1, y=c(.9,1), z=1.1,levels(loadings[[4]]),col="black")
points3d(x=1.2,y=c(.9,1),z=1.1, col=as.numeric(as.factor(loadings[,4])), size=5)

The plot with labels according to classes:

Answer 1

That's not a plot3d image (unless perhaps you have another package loaded) , but looks like a scatter3d rendering constructed with the scatter3d function from the car package by John Fox:

  require(rgl)
  require(car)
  scatter3d(x=loadings[[1]], y=loadings[[2]], z=loadings[[3]], 
            point.col = as.numeric(as.factor(loadings[,4])), size = 10)

The scatter3d function does depend on rgl functions but has quite a few customization options. You offered no code to construct a "legend", so I played around with the example offered in rgl::text3d:

 text3d(1+ font/8, 1+cex/4, 1+famnum/8, text=paste(family, font), adj = 0.5, 
       color="blue", family=family, font=font, cex=cex)

With the new data this is in response to the request for text and points:

 scatter3d(x=loadings[[1]], y=loadings[[2]], z=loadings[[3]], 
             point.col = as.numeric(as.factor(loadings[,4])), size = 10)
 text3d(x=1.1, y=c(.9,1,1.1), z=1.1, names(loadings) ,col="black")
 points3d(x=1.2,y=c(.9,1,1.1),z=1.1, col=as.numeric(as.factor(loadings[,4])), size=5)

Answer 2

I used the formula method to create groups, and 'removed' the group labels by turning them white because I could not figure out how to do this an easier way. Then I created a legend using legend3d ... which was very simple. Just make sure point and legend colours match. Maybe this is clunky, but it did what I needed it to do!

scatter3d(Biomass~Salinity*Moisture|Competitor, data=phrag19, labels=FALSE,
    surface=FALSE,
      grid = FALSE, 
        ellipsoid = FALSE,
    axis.col = c("black", "black", "black"),
    text.col=c("white","white","white","white","white","white","white"),
    surface.col=c("#1D1B1A","#C70039", 
"#94CC00","#A5D7F3","#D2A7DC","#1CAD7A","#FFB114"))

legend3d("right", legend = levels(phrag19$Competitor),
   col =  c("#1D1B1A","#C70039", "#94CC00","#A5D7F3","#D2A7DC","#1CAD7A","#FFB114"), 
pch = 18)

legend3d