0

The quiz I met first gave me an Logistic model:

And ask me to linearize it, then evaluate the value of a and k according to the data it gave( in this subject L is took as 3000). I finished that, but got into trouble in the second subject which asked me to do a non-linear-regression with a and k's value evaluated in the first subject. Here's my code:

function y = func(const, t)
  y = const(1)./(1 + const(2)*exp(-const(3)*t));
end


t = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12];
y = [43.65,   109.86,   187.21,  312.67,   496.58,  707.65 ,  ... 
     960.25,  1238.75,  1560,    1824.29,  2199,    2438.89,  2737.71];

yAss = log ((3000 ./ y) - 1);
p    = polyfit (t, yAss, 1);
a    = exp (1) ^ (p(2)); 
k    = -p(1);

beta0 = [3000, a, k];
beta  = nlinfit (t, yAss, @func, beta0);

yAfter   = beta(1) ./ (1 + beta(2) * exp (-beta(3) * t));
yCompare = 3000 ./ (1 + a * exp (-k * t));

scatter (t, y); hold on;
plot (t, yAfter, 'r');
plot (t, yCompare);

And what it gave:

The red curve is generated with the value returned by nlinfit, Anyone could tell me what is wrong?

Tasos Papastylianou
  • 21,371
  • 2
  • 28
  • 57
Amarth Gûl
  • 1,040
  • 2
  • 14
  • 33

1 Answers1

0

I feel stupid... Answering my own question and the question itself is stupid either...

beta  = nlinfit (t, yAss, @func, beta0);

should be:

beta  = nlinfit (t, y, @func, beta0);

I really want to delete that question...

Amarth Gûl
  • 1,040
  • 2
  • 14
  • 33