Building models with SAS Enterprise Miner, SAS Factory Miner, SAS Visual Data Mining and Machine Learning or just with programming

Capping and Flooring outliers - Methods

Accepted Solution Solved
Reply
Contributor
Posts: 65
Accepted Solution

Capping and Flooring outliers - Methods

Hi,

I am trying to understand and find out what are the different techniques commonly used to Caps and floor outliers present in a dataset? Any guidance/research paper link is greatly appreciated?


Accepted Solutions
Solution
‎09-25-2017 03:47 PM
Super Contributor
Posts: 337

Re: Capping and Flooring outliers - Methods

Hi Pritish,

Take a look at the Replacement node in SAS Enterprise miner. It does not remove outliers the way the Filter node would.

By default the limit method is based on a standard deviation from the mean. Take a look at the flooring and capping that happens by default. There is also a brief example of how to use the replacement editor in the reference help (press F1 and go to the Replacement node under the Modify sections).

Good luck!

Miguel

View solution in original post


All Replies
Super User
Posts: 10,044

Re: Capping and Flooring outliers - Methods

Usually , you can define outliers as   if  >  mean+2*std  or  < mean-2*std

Contributor
Posts: 65

Re: Capping and Flooring outliers - Methods

Thanks! How do you treat those outlier? I don't want them to exclude from my dataset.

Solution
‎09-25-2017 03:47 PM
Super Contributor
Posts: 337

Re: Capping and Flooring outliers - Methods

Hi Pritish,

Take a look at the Replacement node in SAS Enterprise miner. It does not remove outliers the way the Filter node would.

By default the limit method is based on a standard deviation from the mean. Take a look at the flooring and capping that happens by default. There is also a brief example of how to use the replacement editor in the reference help (press F1 and go to the Replacement node under the Modify sections).

Good luck!

Miguel

Super User
Posts: 10,044

Re: Capping and Flooring outliers - Methods

I will Winsorted .  Replace them with 90th or 95th percentile .

☑ This topic is solved.

Need further help from the community? Please ask a new question.

Discussion stats
  • 4 replies
  • 1836 views
  • 0 likes
  • 3 in conversation