Flow Charts with R and the Gmisc package

1 flow charts

1.1 with the Gmisc package

• first create a transition matrix tm, i.e. the rows represent start state the columns represent end state, so tm[1,1] is the number of subjects state in state 1 and end in state 1, so tm[1,3] is the number of subjects state in state 1 and end in state 3
  

tm <- matrix(NA,nrow=3,ncol=3)
tm[1,] <- 200*c(.5,.25,.25)
tm[2,] <- 800*c(.75,.1,.15)
tm[3,] <- 600*c(0,.2,.8)

tm

     [,1] [,2] [,3]
[1,]  100   50   50
[2,]  600   80  120
[3,]    0  120  480

• a more convenient way is to tabulate start state and end state
  

xx <- data.frame(id=1:1000,
                 start.state=sample(1:4,size=1000,replace=T,prob=c(0.2,0.5,0.1,0.2)),
                 end.state=sample(1:4,size=1000,replace=T,prob=c(0.1,0.3,0.2,0.4)))

tm2 <- table(xx$start.state,xx$end.state)
tm2

 
    1   2   3   4
1  24  67  37  89
2  45 143  87 196
3  13  30  24  44
4  25  59  39  78

• than a simple call to transitionPlot() does the magic
  
• box_txt provides the states' names
  

plot.new() ## call plot.new because transitionPlot does not
transitionPlot(tm,box_txt = 1:3)

plot.new() ## call plot.new because transitionPlot does not
transitionPlot(tm2,box_txt = paste("state",1:4))

• of course you can change the appearance of the plot, e.g. color of the boxes, appearance of the arrows
  
• in a similar way text color can be changed
  

plot.new() ## call plot.new because transitionPlot does not

transitionPlot(tm2,box_txt = c(paste("state",1:4)),
               type_of_arrow="gradient",
               fill_start_box=c("midnightblue","darkred","darkgreen","maroon4"),
               fill_end_box=c("midnightblue","darkred","darkgreen","maroon4"))