[R] Problems with nls

[Abolfazl Saghafi]

Can some help me with a question on this bass model, please

As I read some articles on this topic, I understand that
1. the bass formula is
N(t) = pm + (q-p) N(t-1) - (q/m) (N(t-1))^2
2. which is a difference equation with the solution
N(t) = m (1 − exp(−(p+q)t)) / (1 + (q/p)exp(−(p+q)t))
3. So, using a linear regression would give us some some initial
estimations for the parameters m, p, q
4. we then can put the initial estimations into a NLS to get the better
estimations

Am I right?

Now the question is,
why is that I see people use cumulative data and try to fit it into a pdf as
M * ( ((P+Q)^2 / P) * exp(-(P+Q) * T79) ) / (1+(Q/P)*exp(-(P+Q)*T79))^2,

why not using the cumulative data and fit directly the N(t)

	[[alternative HTML version deleted]]