Re: [Gnuplot-bugs] improved interpolation (smoothing)

[<pl...@pi...>]

On Sun, 01 Jul 2007 20:56:11 +0200, Hans-Bernhard Bröker  
<HBB...@t-...> wrote:

> pl...@pi... wrote:
>
>> I think it would be very useful to add lanczos filter as an  
>> interpolation
>> method.
>
> So file it as a feature request.

Well I thought it best to sound the idea out here before taking the time  
to make feature request.
Should I take that sort of reply mean you think it may be a useful idea?


> But make sure you describe the
> background a bit more accurately:

So what is inaccurate?

>
>> The existing splines are fine but have a strong tendancy to extend
>> well outside the path of the existing data when the data has a lot of
>> fluctuations, a common situation.
>
> That's actually only true for csplines and acsplines, but not for any of
> the other filters.

That's why I said ***existing splines*** and not "any of the other  
filters."


>
>> Although the implementation is correct this can be very misleading
>> when viewing data presented with this sort of "smoothing".
>
> And how is that news?  *All* viewing of smoothed data is misleading,
> because it pretends you have smoother data than you actually do.  The
> only question is how the plot misleads the viewer, not if.

What is "news" is that one would expect a smoothing function to reduce the  
excursions from the local mean , not exagerate them. If smoother means  
reduced second differential it's smoother, but badly distorted. I'm not  
sure what real use this sort of smoothing is , it always makes my data  
look worse than the original and gives it a totally misleading form.  
Probably mainly useful for doing pseudo-science with misleading results.


How about dropping the smart, sarcastic tone in *All* your posts? It gets  
rather tiring.