similar to: calculating a gradient in optim without calling a separate function

I use optim to compute maximum likelihood estimations without giving an analytical gradient to optim. However, I would like to get an output of the final numerical gradient vector and the final matrix of contributions to the gradient. But I did not find any mention of this kind of output in help pages. Does anyone know how to do that ? Stephane Luchini GREQAM Marseille, France

Hi all, I would like to use optim() to estimate the equation by the log-likelihood function and gradient function which I had written. I try to use OPG(Out Product of Gradient) to calculate the Hessian matrix since sometime Hessian matrix is difficult to calculate. Thus I want to pick the Gradient matrix from the gradient function. Moreover, could R show the process of calculation on gradient

I have constructed the function mml2 (below) based on the likelihood function described in the minimal latex I have pasted below for anyone who wants to look at it. This function finds parameter estimates for a basic Rasch (IRT) model. Using the function without the gradient, using either nlminb or optim returns the correct parameter estimates and, in the case of optim, the correct standard

Hello All, I am trying to estimate the parameters of a stochastic differential equation (SDE) using quasi-maximum likelihood methods but I am having trouble with the 'optim' function that I am using to optimise the log-likelihood function. After simulating the SDE I generated samples of the simulated data of varying size (I want to see what effect adding more observations has on the

I have had some weird results using the optim() function. I wrote a probit likelihood and wanted to run it with optim() with simulated data. I did not include a gradient at first and found that optim() would not even iterate using BFGS and would only occasionally work using SANN. I programmed in the gradient and it iterates fine but the estimates it returns are wrong. The simulated data work

Hi, I'm very interested in using the constrOptim() function currently in the R-devel sources. In particular, I'm trying to fit point process conditional intensity models via maximum likelihood. However, I noticed that the gradient of the objective function must be supplied for all but the Nelder-Mead method. I was wondering why this was because optim() itself does not require a gradient

Dear Rhelpers I have the problem with initial values, could you please tell me how to solve it? Thank you June > p = summary(maxLik(fr,start=c(0,0,0,1,0,-25,-0.2))) Error in maxRoutine(fn = logLik, grad = grad, hess = hess, start = start, : NA in the initial gradient > p = summary(maxLik(fr,start=c(0,0,0,1,0,-25,-0.2),method="BFGS")) Error in optim(start, func, gr =

> Date: Wed, 20 Mar 2002 14:31:03 +0100 (CET) > From: =?iso-8859-1?Q?G=F6ran_Brostr=F6m?= <gb at stat.umu.se> > Subject: [R] optim with gradient > > I want to maximise a function using 'optim' with a method that requires > the gradient, so I supply two functions, 'fun' for the function value > and 'd.fun' for its gradient. My question is: Since

<FONT face="Default Sans Serif, Verdana, Arial, Helvetica, sans-serif" size=2><DIV>Dear list members,</DIV><DIV>&nbsp;</DIV><DIV>I am trying to use "nlm" function to maximize a mixture likelihood of beta densities. There are five unknown parameters in the likelihood. Since I can get the analytic gradient, I attach the "gradient"

Dear R Users/Experts, I am using a function called logitreg() originally described in MASS (the book 4th Ed.) by Venebles & Ripley, p445. I used the code as provided but made couple of changes to run a 'constrained' logistic regression, I set the method = "L-BFGS-B", set lower/upper values for the variables. Here is the function, logitregVR <- function(x, y, wt =

Sorry, I deleted my old post. Pasting the new query below. Dear R Users/Experts, I am using a function called logitreg() originally described in MASS (the book 4th Ed.) by Venebles & Ripley, p445. I used the code as provided but made couple of changes to run a 'constrained' logistic regression, I set the method = "L-BFGS-B", set lower/upper values for the variables. Here

Dear all, There have been one or two questions posted to the list regarding the optim error "non-finite finite-difference value [4]." The error apparently means that the 4th element of the gradient is non-finite. My question is what part(s) of my program should I fiddle with in an attempt to fix it? Starting values? Something in the log-likelihood itself? Perhaps the data

Dear R help, I am trying solve an MLE convergence problem: I would like to estimate four parameters, p1, p2, mu1, mu2, which relate to the probabilities, P1, P2, P3, of a multinomial (trinomial) distribution. I am using the mle2() function and feeding it a time series dataset composed of four columns: time point, number of successes in category 1, number of successes in category 2, and

Full_Name: Jörg Polzehl Version: 1.5.1 OS: Windows 2000 Submission from: (NULL) (193.175.148.198) When calculating MLE's in a variance component model using constrained optimization, i.e. optim(...,method="L-BFGS-B",...) I observed an inproper behaviour in cases where the likelihood function was evalueted at the constraint. Parameters and value of the function at the constraint

Your function is giving NaN's during the optimization. The R-forge version of optimx() has functionality specifically intended to deal with this. NOTE: the CRAN version does not, and the R-forge version still has some glitches! However, I easily ran the code you supplied by changing optim to optimx in the penultimate line. Here's the final output. KKT condition testing Number of

R-help, I'm just trying to find the ML (maximum likelihood) estimates of the mean and standard deviation of a set of observations: >xx=c(2.5,3.5,4,6,6.5,7.5) fn<-function(params,x=xx) { media<-params[1] st <-params[2] pdf=-sum(dnorm(log(xx),log(media),st,TRUE)) return(pdf) } optim(c(mu,stdev),fn,method="L-BFGS-B",lower=c(0.001, 0.001) ,upper = rep(Inf, 2),

Hello, S-plus and R helpers,(sorry for cross-post) I wrote some simple C code for one likelihood to be optimized (using optim(MASS)). I use same function, same data, same starting points and same DLL in R and S-plus for comparison. (I compiled it with 'Rcmd SHLIB likelihood.c' and the header files of it include only R.h and math.h). While it works quite fine in S-plus, it forever returns

hi Suppose I have a mixture of 2 distributions generated by rtwonormals <- function(npnt,m1,s1,m2,s2,p2){ rv<-vector(npnt,mode="numeric") for( i in seq(1:npnt)){ if(runif(1,0,1)<=p2){ rv[i]<-rnorm(1,m2,s2) } else{ rv[i]<-rnorm(1,m1,s1) } } return(rv) } x <- rtwonormals(50000,0,100,500,500,0.05) #and I try to fit these with (based on thread: [R]

hi , lik <- function(nO, nA, nB, nAB){ loglik <- function(par) { p=par[1] q=par[2] r <- 1 - p - q if (c(p,q,r) > rep(0,3) && c(p,q,r) < rep(1,3) ) { -(2 * nO * log (r) + nA * log (p^2 + 2 * p * r) + nB * log (q^2 + 2 * q * r) + nAB * (log(2) +log(p) +log(q))) } else NA } loglik }

Dear all, I am using the functions 'optim' and 'nlminb'. For both, you can provide a function which computes the gradient of the objective function (to enhance speed and precision). In my case, both the objective function and the gradient take time to be computed and share many common computations (similar matrix, products, etc...). Therefore, I have to compute these