ggplot: ddply(plyr)

1. ddply(.data, .variable,.fun,...)

.data: is the data to break up

.variable: the grouping variable used to break up the dataset.

.fun: summary function you want to use to each group. return a vector or dataframe

use subset() to select qualified subgroups

ddply(diamonds,.(color),subset,carat==min(carat))

use transform()  for transformation

use colwise() to apply a function to selected column in the dataframe

colwise(function): use a function to dataframe, or selected columns of a dataframe

e.g. colwise(mean)(msleep)

numcolwise(): only used for numeric column

catcolwise(): only use for category column

colwise() combined with ddply(), to produce per-group summaries

ddply(msleep, .(vore),numcolwise(median),nam.rm=T)

smooth=function(df){

mod=gam(price~s(carat,bs="cs"),data=df)

grid=data.frame(carat=seq(0.2,2,length=50)

pred=predict(mod,grid,se=T)

grid$price=pred$fit

grid$se=pred$se.fit

grid}

smoothes=ddply(dense,.(color),smooth)

qplot(carat,price,data=smoothes,color=color,geom="line")