I am confused between outliers and leverage points. And the difference between good and bad leverage points in time series analysis.
Can somebody help me?
Asked
Active
Viewed 52 times
0
-
Outliers at the end of time series can have a large impact on forecasting and its very hard to control unless you know a priori via domain knowledge what is going on in your time series, see for example in this post the impact of last data point being outlier https://stats.stackexchange.com/questions/146098/ets-function-how-to-avoid-forecast-not-in-line-with-historical-data/146141#146141 – forecaster Jan 20 '19 at 20:39
-
The terms come from regression. E.g. see here https://stats.stackexchange.com/questions/65912/precise-meaning-of-and-comparison-between-influential-point-high-leverage-point - (a 'good' leverage point is one that would pull the line close to where the bulk of the data would suggest it go (improving precision); a 'bad' leverage point pulls it away from that) – Glen_b Jan 20 '19 at 22:47