#                    Exploring the function acf()

# The function acf(y) will compute the autocorrelation function of the object y, which can
# either be a time series or a vector that is treated like a time series. We'll observe later
# that it can deal with multiple time series, but we will stick with the univaritate case for the moment.

#To have someting to work with, let's just create a random sequence.

y<-rnorm(200)

#Now look at its ACF

yacf<-acf(y)

# Now yacf is the output of the acf function. Let's check its "class"

class(yacf)


#This is an object of type "list".
# How do we ge the names of the items in the list?

names(yacf)

# We see that the names are   "acf"    "lag"    "n.used" "type"   "series" "units"  "call" 

# We can explore these in turn...
# The "bottom line" looks like it might be acf. Let's use the list selector $ to look at it.

yacf$acf

# This looks like an array... lets check

temp=yacf$acf

class(temp)

# Yup, it is an array. We can now pull out the pieces that we want.
# Say we want the first 10 nontrivial correlations in a vector.

v=temp[2:11]

# If you look closely (sometimes a good idea, sometimes not) you will see that temp is # actually a three
# dimensional array. 
# Why? This is because acf is set up to handle multiple series.
# 
# Fortunately we get easy access to the autocorrelations without specifying the "other # dimensions"
# but we could have used the more explicit extraction

v=temp[2:11,1,1]