moving statistics with proc expand

[proctice]

I want a movsum 4 but I don't want that first couple of observations that are computed with 1, 2, 3 observations.  I only want observations that are computed with 4 observations.  I have several by groups so it is not that easy to just delete them. Is there a way to tell proc expand not to include them or an easy way to delete when you have many by groups. Thanks.

[dw_sas · Accepted solution]

Hi,

You can add the TRIMLEFT 3 operand after the MOVSUM 4 specification in your CONVERT statement to set the first 3 moving sum values within each BY group to missing.  For example:

convert x=movsum_x / transform=(movsum 4 trimleft 3);

I hope this helps!

DW