How to identify and remove outliers?

[Babloo]

Could you please help me understand how can I detect and remove outliers? I tried to read the relevant documents but I find difficult to understand.

[ballardw]

The first thing is for you to define clearly to yourself what an "outlier" will be. Typical rules are something like x units of difference from a mean (or median) value. Units of difference might be standard deviations, multiples of the Interquartile range or something else or perhaps the smallest and/or largest x percentage of values.

Graphing data is often one way to see if you have values extreme enough that you think they should be eliminated.

Once you can do that definition then providing code to do such depends on the type of rule and data involved.

[Babloo]

Could you please tell me how to graph the data to find the outliers and how
can I eliminate it?

[ChrisBrooks]

There's quite a nice introduction to outliers in this paper which not only covers what they are and how they arise but gives some simple code examples to help you find them -> https://www.lexjansen.com/nesug/nesug10/ad/ad07.pdf

[Reeza]

Have you taken the free SAS e-course on Statistics with SAS? If not, it's free and available via e-learning.

[StatDave · Accepted solution]

See the "Outlier detection" item in the list of Frequently Asked-for Statistics (see the Important Links section).