Tom Reilly

Waging a war against how to model time series vs fitting

  • Home
    Home This is where you can find all the blog posts throughout the site.
  • Categories
    Categories Displays a list of categories from this blog.
  • Tags
    Tags Displays a list of tags that has been used in the blog.
  • Bloggers
    Bloggers Search for your favorite blogger from this site.

Tell me how you handle outliers in your time series analysis/forecasting process.

Posted by on in Forecasting
  • Font size: Larger Smaller
  • Hits: 51610
  • 0 Comments
  • Subscribe to this entry
  • Print
  • PDF

15 Questions surrounding outliers and time series forecasting/modeling

 

Does your current forecasting process automatically adjust for outliers? (correct answer Yes)

Do you make a separate run for certain problems that SHOULD NOT get adjusted for outliers as the outliers are in fact real and shouldn't be adjusted?  (correct answer Yes)

Do you know what standard deviations are used to identify an outlier?  (correct answer "who cares" You shouldn't be having to tell the system)

Who knows that the standard deviation calculation is itself skewed by the outlier?  (correct answer "who cares" You shouldn't be having to tell the system)

Does the system ask you how many times it should "iterate" when removing outliers? How many times do you "iterate"? (correct answer "who cares" You shouldn't be having to tell the system)

Does the system allow you to convert outliers to causals and flag future values when the event will happen again?  (correct answer Yes)

Does the system identify inliers? ie. 1,9,1,9,1,9,1,5  (correct answer Yes)

Does the system recognize the difference between an outlier and a seasonal pulse?  (correct answer Yes) (IE 1,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,1,0,etc)

Does the system recognize the difference between an outlier and a level shift?  (correct answer Yes) (IE 0,0,0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1,1,1,etc)

Does the system recognize the difference between an outlier and a change in the trend?  (correct answer Yes)  (IE 0,0,0,0,0,0,0,1,2,3,4,5,6,7,8,etc)

Does the system allow you to force the outlier in the most recent period to be a "level shift" or a "seasonal pulse"?  (correct answer Yes)

Does the system report a file adjusted for outliers for pure data cleansing purposes?  (correct answer Yes)

Does the system adjust for outliers in a time series regresion (ie ARIMAX/Transfer Function)? (correct answer Yes)

Who tries to find the assignable cause why the outliers existed?   (correct answer I do)

Who then provides causals to explain the outliers to the system?   (correct answer I do)

Comments

Go to top