[Gretl-devel] Re: What to do when convergence is no reached after 800 iterations (nonlinear regression)

On Thu, 26 Dec 2019, Ivan Falconi wrote: > Hello. > > Please, I need ssome help. > > I am trying to obtain the parameters of the Bass model (p, q , m are > coefficient of innovation, coefficient of imitation and market potential, > respectively) . The message I got was that after 800 iterations convergence > was not reached. > > The initial parameters were obtained with Excel's Solver. I had the same > issue before and I used the program NELREG which gave me same error but I > used a script and was able to find convergence. I am sure the same will > happen with greatl provided I use proper script. This seems to work (with some help from https://srdas.github.io/MLBook/productForecastingBassModel.html ): <hansl> open http://ricardo.ecn.wfu.edu/pub/gretl/apple_sales.gdt # use quarterly iPhone sales data smpl 2007:2 ; series s = iPhone series S = cum(s) series S2 = S^2 # regress sales on cumulated sales and its square ols s const S S2 matrix b = $coeff # unscramble m, p and q m1 = (-b[2] + sqrt(b[2]^2 - 4*b[1]*b[3])) / (2*b[3]) m2 = (-b[2] - sqrt(b[2]^2 - 4*b[1]*b[3])) / (2*b[3]) m = m1 > m2 ? m1 : m2 p = b[1]/m q = -m*b[3] # index starting in 2007Q2 series idx = seq(1,$T)' nls s = (m*((p+q)^2/p)*empqi)/(1+(q/p)*empqi)^2 series empqi = exp(-(p+q)*idx) params m p q end nls </hansl> Allin _______________________________________________ Gretl-devel mailing list -- gretl-devel(a)gretlml.univpm.it To unsubscribe send an email to gretl-devel-leave(a)gretlml.univpm.it Website: https://gretlml.univpm.it/postorius/lists/gretl-devel.gretlml.univpm.it/