sfit object. can't get coefficients.

Hello. There is my code:

>> c = fit( [X, Z], Y, ft, opts );
>> c

       Linear interpolant:
         c(x,y) = piecewise linear surface computed from p
         where x is normalized by mean 0.5 and std 0.2745
         and where y is normalized by mean 227.7 and std 80.16
       Coefficients:

           p = coefficient structure

>> class(c)

ans =

sfit

>> coeffvalues(c)

    ans = 

    TriangleSurfaceInterpolant with properties:

      Type: 'Linear'

>> c.p

ans = 

    TriangleSurfaceInterpolant with properties:

      Type: 'Linear'

>> c(0.3, 201)

ans =

          26349

Is it possible to get coefficient of

c() function?

Answer 1

[Greig]

As Sean de Wolski points out in the question that you linked to, a interpolated surface fit has multiple sets of linear fit coefficients, one for each point in the data.

If you want to have just a single set of coefficients for the entire surface fit you will need to use a polynomial fit. For example, if you want to use a 2nd order polynomial in both x and z, set the fittype to 'poly22'. The c.p00, c.p10, c.p01, c.p11, c.p20, c.p02 will contain all of your coefficients.

If this is not what you are looking for could you gives us some more details about what your are wanting.

[almaz ham]

Greig, thank you for answer. See in the picture. There ar three set of points: PWM, RPS and SOC. I get this picture using result of fit() function. Now i want to get relationship: RPS and SOC ---> PWM. I mean i need to compute PWM value for pair of given SOC and RPS. I need equation to implement it in C code . And this equation should be as easy as possible. But

fit()

function can't give me this equation. Is there any matlab function to fit set of 3 variable? "...use a polynomial fit..." Did you mean

polyfit()

function?

[Greig]

Judging by the pic, you want a linear fit in both x and y, so fit() will give you the answer. What you want is...

c = fit([SOC, RPS], PWM, 'poly11');

As before c.p00, c.p10, c.p01 are the coefficients.

This fits a linear function in both SOC and RPS, that is...

PWM = p00 + p10 * SOC + p01 * RPS

You may want to try

cftool

The interactive fitting tool, that will allow you to play around with the type of fit you use, and it displays the stat

[maitehar]

Hello, reading this because is in the direction of what I want.

What if I want those coeficients,from a linear fit, for each fitted interval?

This is just because, by a linear fit, it simplifies a lot some calculations and allows me to propose an easy analitycal solution in my study.

Any ideas?

[maitehar]

Ah, got it! I'lll leave it here just in case someone checks this later:

The structure with all coeficient values from a linear fit is called:

your_fit_name.p.coefs

cheers,

Answer 2

[almaz ham]

http://www.mathworks.com/matlabcentral/answers/104975-how-i-can-find-the-formula-from-an-linear-interpolation-in-cftool

Answer 3

[almaz ham]

How implement it in embedded code?