Problem using nlinfit . The code works with a random model function that I found on the web, but not with my model functions

Please Help. I have two models that I a trying to fit into a set of data, but I get the following error:

??? Error using ==> nlinfit at 120 Error evaluating model function '@(b,t)(((kb-ki)*(1/exp((t*b(1))/b(2))-1))/b(1))'.

Error in ==> regrassion at 53 b1hat=nlinfit(t,y1,wd,b0); %one volume

Caused by: Error using ==> mldivide Matrix dimensions must agree.

Here is my code:

t=[0, 62 ,121, 150]; y1=[0,.1808318, .2245815, .2469898]; kb=2.2; ki=6.67;

wd=@(b,t) (((kb - ki)*(1/exp((t*b(1))/b(2)) - 1))/b(1)); %first model w1d=@(p,t) -((kb - ki)*((((t*p(... %second model solution is quit long so I am not including it here

b0=[100 5]; p0=[100 200 1 8]; %initial guesses

plot(t,y1,'s','markersize',5,'color',[0,0,0]); hold on

b1hat=nlinfit(t,y1,wd,b0); %one volume b2hat=nlinfit(t,y1,w1d,p0) %two volume

I have a feeling that my model functions are not being understood well. I used the symbolic expression to solve for the model functions and when I graph them separately, they both work. They just don't work when I use them as model function. Please help. I greatly appreciate it.

Answer 1

[Tom Lane]

It looks like you want "1./exp(...)" in place of "1/exp(...)" so you get the reciprocal of each element instead of a matrix division.

Consider each use of *, /, or ^ in your expression to make sure it's as intended. You may find it easier to debug if you just make sure your function evaluates from the command line with a vector of input values, before you try passing it into nlinfit.