The code below shows an example of the animint package: an interactive animation can be defined as a list of ggplots with showSelected and clickSelects aesthetics.

library(animint)
data(breakpoints)
seg.max <- 13
roc <- subset(breakpoints$roc, segments <= seg.max)
some.models <- subset(breakpoints$error, segments <= seg.max)
only.error <- subset(some.models, type=="E")
only.segments <- subset(only.error,samples==samples[1])
signal.colors <- c(estimate="#0adb0a",
                   imprecision="#0098ef")
breakpoints$imprecision$line <- "imprecision"
breakpoints$segments$line <- "estimate"
breakpoints$breaks$line <- "estimate"
viz <- 
  list(signal=ggplot()+
       geom_point(aes(position, signal, showSelected=samples),
                  data=breakpoints$signals)+
       geom_line(aes(position, signal, colour=line),
                 data=breakpoints$imprecision)+
       geom_segment(aes(first.base, mean, xend=last.base, yend=mean,
                        colour=line,
                        showSelected=segments,
                        showSelected2=samples),
                    data=breakpoints$segments, size=3)+
       ggtitle(paste("Noisy signal (black) and model (green)",
                     "with imprecision component of breakpointError (blue)"))+
       geom_vline(aes(xintercept=base, colour=line,
                      showSelected=segments,
                      showSelected2=samples),
                  linetype="dashed",
                  data=breakpoints$breaks)+
       scale_colour_manual(values=signal.colors),
       error=ggplot()+
       geom_line(aes(segments, error, group=type,
                     showSelected=samples,
                     colour=type, linetype=type, size=type),
                 data=some.models)+
       make_tallrect(only.segments, "segments")+
       ggtitle("breakpointError and its components")+
       geom_line(aes(segments, error, group=samples,
                     clickSelects=samples),
                 data=only.error, lwd=4),
       roc=ggplot()+
       geom_path(aes(FPR, TPR, group=samples,
                     clickSelects=samples), data=roc, lwd=4)+
       geom_point(aes(FPR, TPR, showSelected=samples,
                      clickSelects=segments), alpha=3/4,
                  colour=signal.colors[["estimate"]],
                  data=roc, size=5)+
       xlab(paste("False positive rate =",
                  "probability(predict breakpoint|no breakpoint)"))+
       ylab(paste("True positive rate =",
                  "probability(predict breakpoint|breakpoint)"))+
       ggtitle("ROC curves"),
       width=list(signal=1000, error=500, roc=500))
## Make manual scales.
fp.fn <- list(colour=c(FP="skyblue",FN="#E41A1C",E="black",I="black"),
              linetype=c(E="solid",FP="solid",FN="solid",I="dashed"),
              size=c(E=1,FP=3,FN=3,I=1)*2)
for(a in names(fp.fn)){
  values <- fp.fn[[a]]
  labels <-
    c("False negative", "imprecision", "False positive", "breakpointError")
  sc <- ggplot2:::manual_scale(a, values, name="component",
                               breaks=c("FN", "I", "FP", "E"),
                               labels=labels)
  viz$error <- viz$error + sc
  print(viz$err)
}
gg2animint_knitr(viz) # use gg2animint(viz) instead if not inside knitr/Rmd.

• Click on the error or ROC curves to change the number of samples — it should be apparent that when more data are sampled, it is possible to achieve lower breakpointError.
• Click on the background of the breakpointError plot or the green dots on the ROC plot to change the number of segments in the fitted least squares segmentation model.