I know I'm not supposed to use them... but they're just so easy! I
have trouble defining an appropriate function for plyr or apply!
data<-rnorm(144)
groups1<-c('a','b','c','d')
groups2<-c('aa','bb','cc','dd')
machines<-1:12
df<-data.frame(machine=machines,group1=groups1,group2=groups2,U=data,V=2*data,W=data^2,X=1/data,Y=data+2,Z=2/data)
So... I am currently generating a table and a geom_boxplot and squish
em together with gridExtra. But, for columns U,V and W I want to use
group1 as my split variable and columns X, Y and Z I will use group2.
I also need to make it as flexible as possible.
What I've got now is...
box.vars<-match(c('U','V','W'),colnames(df))
index.group<-match('group1',colnames(df))
group.types<-unique(df[,index.group])
for(j in 1:length(group.types)){
for(i in 1:length(box.vars)){
index.rows<-which(df[,index.group]==group.types[j] &
df[,box.vars[i]]!=0)
p<-ggplot(data=df,aes(x=factor(df$machine[index.rows]),y=df[index.rows,box.vars[i]]))
p<-p+geom_boxplot()+labs(x='Machine ID',y=names(df[box.vars[i]]))
p<-p+opts(axis.text.x=theme_text(angle=50,size=7))
mins<-round(tapply(df[index.rows,box.vars[i]],df$machine[index.rows],min),digits=3)
maxes<-round(tapply(df[index.rows,box.vars[i]],df$machine[index.rows],max),digits=3)
medians<-round(tapply(df[index.rows,box.vars[i]],df$machine[index.rows],median),digits=3)
table.out<-data.frame(min=mins,median=medians,max=maxes)
# + misc. gridExtra lines
}
}
Currently I hard code the box.vars and index.group which is ok with
me, but the for loops should be in a fancy function. Anyway, im sure
theres an elegant plyr or apply that can do this for me... but as I
said before, I need a FA Group (for loops anonymous)...
Also, this winds up being a lot of calcs on a big data set. So, if
you have magical ff, big.memory and/or doMC suggestions I'm all ears,
I just have very little understanding of how they're working.
Thanks for your help,
Justin
