Time series of the day
Stat Sept 2008 D. R. Brillinger Simple descriptive techniques Trend X t = + t + t Filtering y t = r=-q s a r x t-r Simple moving average s = q, a r = 1/(2q+1) Filters may be in series
Differencing y t = x t - x t-1 = x t "removes" linear trend Seasonal variation model X t = m t + S t + t S t S t-s 12 x t = x t - x t-12, t in months
Stationary case, autocorrelation estimate at lag k, r k t=1 N-k (x t - )(x t+k - ) over t=1 N (x t - )2 autocovariance estimate at lag k, c k t=1 N-k (x t - )(x t+k - ) / N
Departures from assumptions Nonstationarity Trend - OLS Seasonality - trig functions Missing values Outliers