STATS 499 Lecture Notes - Lecture 21: Statistical Inference, Autocorrelation, Time Series
Document Summary
Vanish quickly after 5 or 6 lags. Figure 1 : 1st autocorrelation is positive. Figure 2: 1st autocorrelation is negative (the plot shown in fig. If the first autocorrelation is negative, then we would expect alternating signs at the other lags: 2) seasonality. Example of walmart quarterly sales: look for patterned spike in graph. In the least squares regression: = 2 ; = 2 2 = In time series data, we need to take care of : trend, seasonality, then come up with a model to forecast, example of time series of airline data from 1949 to 1960 , a) original data set. From the graph, there is an upward trend need to remove the linear trend first. Constant variance violated too use log transformation: c) using log function and added month indicator. Red dots are based on the model.
