R help - Jul 2011 - How to predict significant dominant regions of two sequence of numeric values by Hidden Markov Model

Dear all,

I have few training on Hidden Markov Model. But, I intend to solve my
problem by HMM. I would like to have your helps/directions to me.

Here, I have two variables to define the 8 one-dimension space (coordinate.1,
coordinate.2). In this one-dimension space, there are two sequences of
values (shared and specific). This means I would like to
detect/guess/predict the regions (defined by two coordinates variables)
which are significantly dominant (with higher values) by shared/specific in
some consecutive cells (units of the coordinate.2 in the dummy). I would
like to get the coordinates (the 1st two variables) of such dominant regions
for shared or specific.

Thanks in advance.

Best regards,

Jian-Feng,

########################################################

I use R to construct a dummy (the real data are more complex than this

Here is my data set:

mydata <- data.frame(coordinate.1=rep(1:8, each=25),
                     coordinate.2=rep(seq(100, 2500, 100), 8),
                     shared=rep(c(100,30,100), c(5,15,5)),
                     specific=rep(c(25,90,20,30), c(5,7,8,5)))

Here is how I happen to make a plot:

library(ggplot2)

pdf("shared_specific.pdf", width = 14, height = 8)
p.test<-ggplot(mydata, aes(coordinate.2)) +
       geom_line(aes(y = shared, colour = "shared")) +
       geom_line(aes(y = specific, colour = "specific")) +
       facet_grid(coordinate.1 ~., scales = "free_x") +
       scale_x_continuous("coordinate.2") +
       scale_y_continuous("shared and specific")
p.test
dev.off()

	[[alternative HTML version deleted]]