[Scilab-users] {EXT} Re: Multiple regression on semi-log plot

[Stéphane Mottelet]

Hi,

Le 16/11/2020 à 10:26, Dang Ngoc Chan, Christophe a écrit :
> Hello,
>
>> De : users <users-bounces at lists.scilab.org> De la part de Stéphane Mottelet
>> Envoyé : lundi 16 novembre 2020 09:56
>>
>> It is possible to fit a more general piecewise-affine model,
>> but you will have to know in advance the number of "kinks"
> Well, I think it is possible to automatically adjust the number of segments:
>
> 1. Perform the regression with only n = 1 line segment.
>
> 2. Increase the number of segments by 1, n = n + 1.
> Perform a new regression.
> You can detect the furthest point to know where to separate the segments.
>
> 3. Compare the quadratic errors between the current and the preceding regression.
> If the difference is small enough then end

That's the more difficult part. Considering the provided data, which 
will never fit well this kind of model, the quadratic error will 
decrease quite regularly. If the real concern is about finding the 
points with maximum curvature, a non parametric model should be more 
adequate (Local regression, Kernel regression, splines, or wathever).

S.


> else go back to 2.
>
> Regards
>
> --
> Christophe Dang Ngoc Chan
> Mechanical calculation engineer
>
> General
> This e-mail may contain confidential and/or privileged information. If you are not the intended recipient (or have received this e-mail in error), please notify the sender immediately and destroy this e-mail. Any unauthorized copying, disclosure or distribution of the material in this e-mail is strictly forbidden.
> _______________________________________________
> users mailing list
> users at lists.scilab.org
> https://antispam.utc.fr/proxy/1/c3RlcGhhbmUubW90dGVsZXRAdXRjLmZy/lists.scilab.org/mailman/listinfo/users

-- 
Stéphane Mottelet
Ingénieur de recherche
EA 4297 Transformations Intégrées de la Matière Renouvelable
Département Génie des Procédés Industriels
Sorbonne Universités - Université de Technologie de Compiègne
CS 60319, 60203 Compiègne cedex
Tel : +33(0)344234688
http://www.utc.fr/~mottelet