Smoothing data

[ogri]

Hi

I have a need to take some data and produce a smoothed set of data containing the inbetween values. Rather like a smoothed graph, except I need the values to do some calculations on rather than just display them.

Any common routines I can use to pass a small subset of data and the call to return values (calculated if not actually present in the original data).

All the best

Keith

 

[traingamer]

Could you clarify that a little? Data example, maybe - with input and expected output?

Greg
"Personally, I am always ready to learn, although I do not always like being taught." - Winston Churchill

 

[ogri]

Hi

Probably best to tell you what I am trying to do.

Basically it is for doing some calculations on an engines power output at certain rpm. But from a limited number of points.

eg, have the power known at 1500rpm, 4000rpm, 5000rpm, 6000rpm, 6500rpm, and from that want to produce inbetween figures (say, for every 100rpm). At the moment it is done crudely with a straight line between each known point.

Thanks.

All the best

Keith

 

[MichaelRed]

there are numerous 'curve fitting' routines in the public domain, but i haven't seen any in this (MS Access orientated) forums. one set of procedures i have used (showing my age her) is called "linear" developed by / at Oak Ridge Research Labirtories in the 1960's. They were written in FORTRAN ver II - I think) and are available in the public domain. If you search the internet, you should easily find references, some of which will have hte entire source code (in FORTRAN - of course!) and could perhaps translate them to VB(A). don't be put off by the name "Linear", it does not refer to the "interpolation", but rather the technique "linear regression" where there is the potential to obtain a least squares fit of the supplied data over several different 'degrees' of the equation form and the option of generating the data for several different equation forms.

Another approach you could try is refered to as a "spline fit", which attempts to generate a set of equations representing different segments (groups) of the data points and to weight an overlap of the various segments t produce a smoth final "curve" of the overall data set.

And, Finally, EXCE: -for all of its faults- does include a regression analysis which may be used for this purpose.

Your profile here does not provide much insight into your background / capability, so you will need to decide which (if any) of these are reasonable "fits".






MichaelRed

 

[ogri]

Hi

Cheers for that. Now I know what to look for. Hopefully can take the Fortran code and work something out.

Background wise, been coding for about 20 year, in pl/1, Cobol, VB and a few other nasties.

All the best

Keith