Put the leverage and outlier observations in a separate dataset

[fatemeh]

Hello all,

I want to put the leverage and outlier observations that are detected via robust regression, in a separate dataset. Any help will be appreciated to solve my problem.

[mkeintz · Accepted solution]

@fatemeh 

In your recent thread Detect extreme values through leverage and outlier changing cutoff value based on 90 percentiles you were recommended to read @Rick_SAS's blog post Detecting outliers in SAS: Part 3: Multivariate location and scatter. 

Take another look at the post.  It has an example that does bulk of what you request.  The statement 

   ods output diagnostics=Diagnostics(where=(leverage=1));

produced a data set named diagnostics with 4 observations at leverage=1. 

Instead you could drop the "where=(leverage=1)" options, which would produce as many obs in diagnostics as in your original data, and in the same order.

You could then do this:

data want;
   merge your_original_dataset diagnostics;
   if leverage=1;
run;

Or you could use whatever minimum leverage values you prefer.