Re: Godin Filter

[kessler@xxxxxxxxxxxxx (Billy Kessler)]

> I am wanting to pass hourly current data through a Godin filter.  That is,
> perform 3 UNWEIGHTED running means - 24, 24, 25 - but am not having much
> luck.  Does anyone our there have an existing journal file or alternatively
> some ideas on how to do it?  

You can use @SBX to make running averages, including those with an even
number of points, but I assume you knew that. If the problem is that 
@SBX:24 weights the endpoints by 1/2 (in order to keep the running mean 
on the original gridpoints), then you can construct such a mean by hand. 
Since you will be doing the even-numbered running mean twice, the result 
will still end up on the original gridpoints:

let rm1=(var[l=@shf:-12]+var[l=@shf:-11]+ ... +[var[l=@shf:11])/24
let rm2=(rm1[l=@shf:-11]+rm1[l=@shf:-10]+ ... +[rm1[l=@shf:12])/24
let godin=rm2[l=@sbx:25]

In this example, the result of the first calculation (rm1) will be placed
"incorrectly", in that the first value will be found at point 13, whereas
a true 24-point running mean would be centered at point 12.5. But then the
second calculation (rm2) will place its first result at gridpoint 24 (note
that the arguments to @SHF are different in each case), which is correct.
Then then final filter (godin) is a straightforward @SBX, which will place 
the first value at gridpoint 36.

You will probably run into line-length problems with the specification of
rm1 and rm2. In that case break the calculation into a few segments.

I'm curious if the result of this is significantly different than what you
would get by simply doing:

let check1=var[l=@sbx:24]
let check2=check1[l=@sbx:24]
let check3=check2[l=@sbx:25]

????

Billy K