This fits to my findings so far.
A tip from a friend:
Use the user programming feature and define it to one of the user definable keys (as I did for the TEMP? command).
But we did not find the solutiion for the rolling average, just the one time average (STATS, MEAN).
My solution so far:
set the NPLC to 10*N, according to the manual, in case NPLC is > 10, the tools does N x (NPLC=10 measures) and builds the average.
-quote-
if you specify 60=PLCs (1 second of integration time at a 60 Hz = line frequency) using the NPLC command, the mulitmeter averages six 10 PLC readings.
-unquote-
But this is still no rolling average unfortunately.