multi Exponential regression method

I work regression about a number of data. so I use matlab fuction : fit(x,y,exp2) but, I wonder that How do get a, b, c, d of f(x)=a*exp(bx)+c*exp(dx).

Answer 1

[Torsten]

If you use

f=fit(x,y,exp2)

f is automatically shown after execution.

Best wishes

Torsten.

Answer 2

[John D'Errico]

For example:

mdl = fit(rand(100,1),rand(100,1),'exp2')
mdl = 
     General model Exp2:
     mdl(x) = a*exp(b*x) + c*exp(d*x)
     Coefficients (with 95% confidence bounds):
       a =       -1101  (-9.55e+11, 9.55e+11)
       b =      0.9633  (-1.213e+05, 1.213e+05)
       c =        1102  (-9.55e+11, 9.55e+11)
       d =       0.963  (-1.213e+05, 1.213e+05)

The coefficients themselves in this case are total garbage, because my data was just random numbers. The massively wide confidence bounds reflect that fact.

Regardless, you can see the values there. But they are printed only to a few significant digits. NEVER just take those numbers as is and try to use them. You need the full values of the coefficients, if you will use them. So, how do you extract the values themselves as numbers?

When you have no idea what applies to a class, then look at methods. It will tell you. First, we can see that mdl is an object of class cfit.

whos mdl
  Name      Size            Bytes  Class    Attributes
  mdl       1x1              1198  cfit

What methods apply to that class?

methods(mdl)
Methods for class cfit:

argnames       coeffnames     dependnames    fitoptions     integrate      numcoeffs      probnames      type           
category       coeffvalues    differentiate  formula        islinear       plot           probvalues     
cfit           confint        feval          indepnames     numargs        predint        setoptions

So, clearly coeffvalues will be of use here. If you are not sure, then use the help:

help cfit/coeffvalues
 coeffvalues Coefficient values.
    coeffvalues(FUN) returns the values of the coefficients of the
    CFIT object FUN as a row vector.

      See also cfit/probvalues, fittype/formula.

YEP. coeffvalues is what you need.

format long g
coeffvalues(mdl)
ans =
   -1101.27586379757  0.963250512625987  1101.7626267532  0.96297061650439

When you have a problem, the help is there to serve your needs. Use it.