Clustering and interactive graph

graph.rmd

Interactive Graph Example
=========================

Function to do the work.  Aside from the packages, to run this, just load the knitr package, and then knit2html("graph.rmd").

```{r}
longplot <- function(dat, k, xlab, ylab, seed=4324, miter = 20) {
require(EMCluster)
require(mice)
require(reshape2)
require(ggplot2)
require(rCharts)

  dat <- as.data.frame(dat)
  set.seed(seed)
  if (any(is.na(dat))) {
    x <- complete(mice(as.matrix(dat), m = 1, method = "pmm", printFlag = FALSE))
  } else {
    x <- as.matrix(dat)
  }
  mm <- init.EM(scale(x), nclass = k, method = "em.EM", min.n.iter = miter)
  dat$id <- factor(1:nrow(dat))
  dat$cluster <- with(mm, factor(class, levels = 1:k, labels = sprintf("%d: n = %d", 1:k, nc)))
  ldat <- na.omit(melt(dat, id.vars = c("id", "cluster")))
  ldat$variable <- as.numeric(factor(ldat$variable, levels = colnames(dat)))
  p <- ggplot(ldat, aes(variable, value, group = id, colour = cluster)) +
    geom_line() + facet_wrap(~cluster) + labs(x = xlab, y = ylab)

  res <- do.call(rbind, by(ldat, ldat$cluster, function(d) {
    out <- data.frame(Y = as.vector(tapply(d$value, d$variable, mean, na.rm=TRUE)),
      X = unique(d$variable))
    colnames(out) <- c(ylab, xlab)
    out$cluster <- unique(d$cluster)
    return(out)
  }))

  return(list(Means = res, plot = p, ldat = ldat))
}
```

Get some public simulated data, set things up, and make the plot.

```{r}
wdat <- read.csv("http://joshuawiley.com/files/simwdat.csv")

require(rCharts)
options(RCHART_WIDTH = 850, RCHART_HEIGHT = 400)


m <- longplot(wdat[, c("prc1T3", "prc1T4", "prc1T5", "prc1T6")], k = 9, "Visit", "PerceivedRisk")

p <- nPlot(PerceivedRisk ~ Visit, group = "cluster", data = m$Means, type = "lineChart")
```

Print the plot

```{r results='asis', comment=NA}
p$print('TestChart', include_assets = TRUE)
```

Individual growth curve plot, by cluster.

```{r fig.width=14, fig.height=10}
m$plot
```

Fin!