Performing some automated forecasting using the ets function

So far, we have looked at ways in which we can explore any linear trends which may be inherent in our data. That provided a solid foundation for the next step, prediction. Now we will begin to look at how we can perform some actual forecasting.

Converting the dataframe to a time series object

As a preparation step, we will use the ts function to convert our dataframe to a time series object. It is important that the time series be equally spaced before converting to a ts object. At a minimum, you supply the time series variable, and start and end dates as arguments to the ts function.

After creating a new object, x, run a str() function to verify that all of the 14 time series from 1999 to 2012 have been created:

# only extract the 'ALL' timeseries
x <- ts(x2$Not.Covered.Pct[1:14], start = c(1999), end = c(2012), frequency = 1)

str(x) 
>  Time-Series [1:14] from 1999 to 2012: 0.154 0.157 0.163 0c.161 0.149 ...