similar to: Entropy/KL-Distance question

Hi, This is more of a general stat question. I am looking for a easily computable measure of a distance between two empirical distributions. Say I have two samples x and y drawn from X and Y. I want to compute a statistics rho(x,y) which is zero if X = Y and grows as X and Y become less similar. Kullback-Leibler distance is the most "official" choice, however it needs estimation of

I am trying to calculate a Kullback-Leibler divergence from two vectors with integers but get NA as a result when trying to calulate the measure. Why? x <- cbind(stuff$X, morestuff$X) x[1:5,] [,1] [,2] [1,] 293 938 [2,] 293 942 [3,] 297 949 [4,] 290 956 [5,] 294 959 KLdiv(x) [,1] [,2] [1,] 0 NA [2,] NA 0 Best, Ralf

I am performing some analysis over a large data frame and would like to conduct repeated analysis over grouped-up subsets. How can I do that? Here some example code for clarification: require("flexmix") # for Kullback-Leibler divergence n <- 23 groups <- c(1,2,3) mydata <- data.frame( sequence=c(1:n), data1=c(rnorm(n)), data2=c(rnorm(n)), group=rep(sample(groups, n,

Dear R Users, I am trying to suppress the information printed by the ecdf function during an assignment. Various alternatives have failed me so far: > a=summary(ecdf(rnorm(100)))["1st Qu."] Empirical CDF: 100 unique values with summary > invisible(a=summary(ecdf(rnorm(100)))["1st Qu."]) Empirical CDF: 100 unique values with summary > (function()

Hey, I want to draw several plots sequently, but have to make them dispaly in one figure. So how to achieve this? Thanks. Fred

Hi there, I'm trying to find the KL divergence measure between a prior and it's posterior distributions, and I'm using the KLdiv method in the flexmix package. plese see the example below: require(flexmix) x=seq(-4,4,length=100) d1=dnorm(x,0,1) d2=dunif(x,-3,3) y=cbind(d1,d2) kl=KLdiv(y) but let say, x1=seq(-5,5,length=100) d3=dunif(x1,-3,3) y1=cbind(d1,d3) kl1=KLdiv(y1) Notice

Sorry, forgot to send this to the list originally. -----Original Message----- From: Mike Waters [mailto:dr.mike at ntlworld.com] Sent: 06 May 2005 18:40 To: 'Campbell' Subject: RE: [R] distance between distributions -----Original Message----- From: r-help-bounces at stat.math.ethz.ch [mailto:r-help-bounces at stat.math.ethz.ch] On Behalf Of Campbell Sent: 06 May 2005 11:19 To:

Hi, I have a vector or random variables and I'm estimating the density using "bkde" function in the KernSmooth package. The out put contains two vectors (x and y), and the R documentation calls y as the density estimates, but my y-values are not exact density etstimates (since these are numbers larger than 1)! what is y here? Is it possible to get the true estimated density at each

Hi, I have come to my wits end again (lately, it's a very short trip). I have been trying to connect to Samba 3.0.21b but Windows keeps throwing up the logon dialog. I'm using "Map network drive" to try to mount this share. It seems the problem is only on the Windows side since I have tested the connection through the Unix account. Both user id and password work fine.

Een ingesloten tekst met niet-gespecificeerde tekenset is van het bericht gescrubt ... Naam: niet beschikbaar Url: https://stat.ethz.ch/pipermail/r-help/attachments/20070803/79b6292b/attachment.pl

Hello, I'd like to say that it's clear when an independent variable can be ruled out generally speaking; on the other hand in R's AIC with bbmle, if one finds a better AIC value for a model without the given independent variable, versus the same model with, can we say that the independent variable is not likely to be significant(in the ordinary sense!)? That is, having made a lot of

A new version of gss, version 0.8-2, is on CRAN now. Numerous new functionalities have been added since my last r-announce post. An ssanova1 suite has been added since version 0.7-4. It implements low-dimensional approximations of the smoothing spline ANOVA models of the ssanova suite. ssanova1 scales much better than ssanova with large sample sizes. A gssanova1 suite is added for non

A new version of gss, version 0.8-2, is on CRAN now. Numerous new functionalities have been added since my last r-announce post. An ssanova1 suite has been added since version 0.7-4. It implements low-dimensional approximations of the smoothing spline ANOVA models of the ssanova suite. ssanova1 scales much better than ssanova with large sample sizes. A gssanova1 suite is added for non

Dear R users, I used sm.density function in the sm package and kde2d() in the MASS package to estimate the bivariate density. Then I calculated the Kullback leibler divergence meassure between a distribution and the each of the estimated densities, but the asnwers are different. Is there any difference between the kde2d and sm.density estimates? if there is a difference, then which is the best

Dear R-Users, my objective is to measure the overlap/divergence of two probability density functions, p1(x) and p2(x). One could apply the chi-square test or determine the potential mixture components and then compare the respective means and sigmas. But I was rather looking for a simple measure of similarity. Therefore, I used the concept of 'intrinsic discrepancy' which is defined as:

Dear R users, According Brockwell & Davis (1991, Section 9.3, p.304), the penalty term for computing the AIC criteria is "p+q+1" in the context of a zero-mean ARMA(p,q) time series model. They arrived at this criterion (with this particular penalty term) estimating the Kullback-Leibler discrepancy index. In practice, the user usually chooses the model whose estimated index is

Hello all, I would like to plot the emperical CDF, normal CDF and pareto CDF in the same graph and I amusing the following codes. "z" is a vector and I just need the part when z between 1.6 and 3. plot(ecdf(z), do.points=FALSE, verticals=TRUE, xlim=c(1.6,3),ylim=c(1-sum(z>1.6)/length(z), 1)) x <- seq(1.6, 3, 0.1) lines(x,pgpd(x, 1.544,0.4373,-0.2398), col="red") y

Hi R-users,   I have this code below and I understand the error message but do not know how to correct it.  My question is how do I get rid of “with absolute error < 7.5e-06” attach to value of cdf so that I can carry out the calculation.   integrand <- function(z) { alp  <- 2.0165   rho  <- 0.868   # simplified expressions   a      <- alp-0.5   c1     <-

Hello! I have a problem with integrate() in my function nctspa(). Integrate produces an error message "evaluation of function gave a result of wrong length". I don't know what that means. Could anyone suggest me what is wrong with my function? These are the examples of function calls that work OK: nctspa(a=1:10,n=5) nctspa(a=1:10, n=5, mu=2, theta=3, renorm=0) This does not work:

My machine has 8GB memory. I had quit all other programs that might take a lot of memory when I try the script (before I post the first message in this thread). The cdf file is of only 741 MB. It is strange to me to see the error. On Mon, Dec 28, 2009 at 2:38 AM, Wolfgang Huber <whuber at embl.de> wrote: > Dear Peng Yu > > how big is the RAM of your computer? You could try with