Help me to make best fit with error

  • Thread starter shad0w2000
  • Start date

Main Question or Discussion Point

Hi,

I have a set of datapoints (x_i,y_i) and I am gonna make a best fit to y = c*Exp(a/x)

Making the fit isn't the problem (Mathematica, etc. can handle this), but what I need is the error on a.

What I have done so far is something like this:

P(c,a | {x_i,y_i} ) = k * P( {x_i,y_i} | c,a ) using Bayes.

And then

P({(x_1,y_1),(x_2,y_2),...}|c,a) = product of ( P(x_i,y_i | c,a) ) for all i's.

which is proportional with

product of ( P(y_i | x_i, c,a) )



But I can't get any further than this :)

Can any one help me with this ?

Thanks in advance
 

Answers and Replies

ssd
268
6
How are you defining the error on 'a'? Totally missing what you mean!
 
EnumaElish
Science Advisor
Homework Helper
2,285
123
y = c*Exp(a/x)
Log y = Log c + a (1/x)
v = d + a z where v = Log y, d = Log c and z = 1/x.

Now you have a linear regression that you can compute with any type of software that will print out the standard errors for the regression parameters d and a. For example, Excel. Or use the Regress function in Math'ca (after you enter Needs["LinearRegression`"]).
 
Last edited:

Related Threads for: Help me to make best fit with error

Replies
4
Views
6K
  • Last Post
Replies
3
Views
4K
Replies
5
Views
2K
  • Last Post
Replies
13
Views
3K
Replies
11
Views
9K
Top