similar to: problem of "constrOptim.nl", no hessian and convergence

Dear Helper, I used "constrOptim.nl" and got the value of par. The estimations looks good even if the number of iterations is only 16. But the values of hessian and convergence are both "NULL". I tested the objective function and gradient function by "optim" and didn't see any problem there. With these functions, "optim" gives the convergence value

Hi, Please see the attached contour plot (I am sorry about the big file). This was created using the vis.gam() function in "mgcv" package. However, my question is somewhat broader. In generating this figure, I first created the contours using vis.gam() and then I plotted the points. These point are plotted on top of the contours so that some of the contour lines are only partially

Hi Reuben, I am puzzled to note that the gradient check in "optimx" does not work for you. Can you send me a reproducible example so that I can figure this out? John - I think the best solution for now is to issue a "warning" rather than an error message, when the numerical gradient is not sufficiently close to the user-specified gradient. Best, Ravi.

Hi, How can I get information on how many times a particular package has been downloaded from CRAN? Thanks, Ravi. ------------------------------------------------------- Ravi Varadhan, Ph.D. Assistant Professor, Division of Geriatric Medicine and Gerontology School of Medicine Johns Hopkins University Ph. (410) 502-2619 email: rvaradhan@jhmi.edu<mailto:rvaradhan@jhmi.edu> [[alternative

Try this: Age_log_model = glm(Arthrose ~ Alter, data=x, start=c(-1, 0), family=quasibinomial(link = log)) Ravi Ravi Varadhan, Ph.D. Assistant Professor The Center on Aging and Health Division of Geriatric Medicine & Gerontology Johns Hopkins University rvaradhan@jhmi.edu<mailto:rvaradhan@jhmi.edu> 410-502-2619 [[alternative HTML version deleted]]

There may be benefits to having a machine learning method that explicitly targets covariate balance. We have experimented with optimizing the weights directly to obtain the best covariate balance, but got some strange solutions for simple cases that made us wary of such methods. Machine learning methods that yield calibrated probability estimates should do well (e.g. those that optimize the

Hi, I am trying to do the following computation: p <- rep(0, n) coef <- runif(K+1) U <- matrix(runif(n*(2*K+1)), n, 2*K+1) for (i in 0:K){ for (j in 0:K){ p <- p + coef[i+1]* coef[j+1] * U[,i+j+1] } } I would appreciate any suggestions on how to perform this computation efficiently without the "for" loops? Thank

Hi, Currently, the "base" has colSums, colMeans. It seems that it would be useful to extend this to also include colMin, colMax (of course, rowMin and rowMax, as well) in order to facilitate faster computations for large vectors (compared to using apply). Has this been considered before? Please forgive me if this has already been discussed before. Thanks, Ravi Ravi Varadhan, Ph.D.

Hi, I would like to know whether there exist algorithms to compute the coefficients or, at least, the degree of the minimal polynomial of a square matrix A (over the field of complex numbers)? I don't know whether this would require symbolic computation. If not, has any of the algorithms been implemented in R? Thanks very much, Ravi. P.S. Just for the sake of completeness, a

Hi, I have a simple problem that I would appreciate getting some tips. I am using the "truehist" function within an "apply" call to plot multiple histograms. I can't figure out how to get truehist to use the column names of the matrix as the labels for the x-axis of the histograms. Here is a simple example: X <- matrix(runif(4000),ncol=4) colnames(X)

Dear R group, I have a data frame which contains data on preferences on 7 items (ranks 1 through 7) listed by each participant. I would like to tabulate this in a 7x7 table where the rows would be the items and the columns would be the number of times that item received a particular rank. I tried doing this by creating a matrix by "rbind"ing each vector obtained using

Hi, I am generating large Kac matrices (also known as Clement matrix). This a tridiagonal matrix. I was wondering whether there is a vectorized solution that avoids the `for' loops to the following code: n <- 1000 Kacmat <- matrix(0, n+1, n+1) for (i in 1:n) Kacmat[i, i+1] <- n - i + 1 for (i in 2:(n+1)) Kacmat[i, i-1] <- i-1 The above code is fast, but I am curious about

Hi, The `system.time(expr)' command provide 3 different times for evaluating the expression `expr'; the first two are user and system CPUs and the third one is total elapsed time. Suppose I want to compare two different computational procedures for performing the same task, which component of `system.time' is most meaningful in the sense that it most accurately reflects the

Hi, Where can I find the source code for nlm()? I dowloaded the R2.8.1.tar.gz file and looked at all the .c and .f files, but couldn't find either nlm.c or nlm.f There is an nlm.r file, but that is not useful. Thanks for any help, Ravi. ---------------------------------------------------------------------------- ------- Ravi Varadhan, Ph.D. Assistant Professor, The Center on Aging

Hi, I have a right-censored (positive) random variable (e.g. failure times subject to right censoring) that is observed for N subjects: Y_i, I = 1, 2, ..., N. Note that Y_i = min(T_i, C_i), where T_i is the true failure time and C_i is the censored time. Let us assume that C_i is independent of T_i. Now, I would like to generate another random variable U_i, I = 1, 2, ..., N, which is

Hi, I am testing code that is working successfully using constrOptim. I am replacing constrOptim with other solvers to compare benefits and performance. My belief was that, like constrOptim, auglag in the alabama package would eliminate beta vectors that failed the constraints prior to calling the objective function. These betas are not only not in the solution space but also cause

Hi, We (Paul Gilbert and I) have just released a new R package on CRAN called "BB" (stands for Barzilai-Borwein) that provides functionality for solving large-scale (and small-scale) nonlinear system of equations. Until now, R didn't have any functionality for solving nonlinear systems. We hope that this package fills that need. We also have an implementation of the

Hi, I am plotting values of log(hazards ratio) as a function of two predictors, using the plotting function filled.contour(). Here is a simple simulated example of this: x <- seq(0,1, length=20) y <- seq(0,1, length=20) z <- outer(x,y, function(x,y) x^2 + y^2 ) zmat <- matrix( rexp(n=400, rate = z+0.001), 20, 20) filled.contour(x, y, log(zmat),

Hi, I would like to know how to capture the console output from running an algorithm for further analysis. I can capture this using capture.output() but that yields a character vector. I would like to extract the actual numeric values. Here is an example of what I am trying to do. fr <- function(x) { ## Rosenbrock Banana function on.exit(print(f)) x1 <- x[1] x2 <- x[2]

Thanks for your message! Actually it works quite well for me too. If I then take the trace of the final result below, I end up with a number made up of both a real and an imaginary part. This does not probably mean much if the trace of the matrix below givens me info about the degrees of freedom of a model... Simona >----Messaggio originale---- >Da: RVaradhan at jhmi.edu >Data: