Hi, There are two kinds of transformations, (try the command "show transformations" for a list of all of them).  Variable transformations return their result on the same underlying axis as the original variable but do some operation such as smoothing, along that axis.  Regridding transformations put the result onto a new axis.  It sounds as if you want a regridding operation.  You need to define a new axis with the desired output time interval and regrid onto that with the @AVE regridding transformation.  That might look like this, if we're just starting with an axis that is t=1,2,3,...,84 yes? DEFINE AXIS/t=2:84:3 t3axis yes? LET p3 = p[gt=t3axis@AVE] Then axis t3axis has points 2,5,8, ..., 83 and the first value of variable p3 is the average of p at 1, 2, and 3 and so on.  If your axis is defined in terms of dates you would do something similar. You might find this FAQ useful: How can I determine the averaging weights when regridding with @AVE? -Ansley On 10/19/2011 10:24 AM, nuncio@xxxxxxxxx wrote: ```Hi, THis must be a naive question. I have a time axis running from 1:84 and an average of every three values need to be computed(not a running mean but mean of first 3 values then second 3 values so on). I tried with a few ways like t=1:84:3@ave, changing the axis to real time axis like one starting from 1-jan-1900. The problem is FERRET computes the average but the first value is the average of the first two values not the first three values @sum also gives sum of first two values with @ngd showing 2 for the first block. here are few outputs yes? sh grid p GRID GIN1 name axis # pts start end normal X normal Y normal Z TAX1 T 84 r 1 84 is the interval is missing here? list/l=1:6 p VARIABLE : P FILENAME : p.cdf SUBSET : 6 points (T) 1 / 1: 0.2661 2 / 2: 0.3101 3 / 3: 0.3174 4 / 4: -0.3617 5 / 5: -0.1323 6 / 6: -0.0020 list/l=1:2 p[t=1:84:3@ave] VARIABLE : P regrid: 3 delta on T@AVE FILENAME : p.cdf SUBSET : 2 points (T) 1 / 1: 0.2881 4 / 2: -0.0589 here the fist value is the sum of the first two values 0.2661 and 0.3101 output of @ngd is this list/l=1:2 p[t=1:84:3@ngd] VARIABLE : P (# of points) regrid: 3 delta on T@NGD FILENAME : p.cdf SUBSET : 2 points (T) 1 / 1: 2.000 4 / 2: 3.000 Why the FERRET is not taking first three values for the first average. Nuncio Murukesh, Ph.D Research Scientist. National Centre for Antarctic and Ocean Research Headland Sada, Vasco Da Gama Goa, 403804 India. Ph: Cell-91 9890357423 off- 91 832 2525428 ```