search for: sum_j

...amer-von Mises type test statistic which involves double sums of max(X_i,Y_j) where X and Y are vectors of differing length. I am currently using ifelse pointwise in a vector, but have a nagging suspicion that there is a more efficient way to do this. Basically, I require three sums: sum1: \sum_i\sum_j max(X_i,X_j) sum2: \sum_i\sum_j max(Y_i,Y_j) sum3: \sum_i\sum_j max(X_i,Y_j) Here is my current implementation - any pointers to more efficient computation greatly appreciated. nx <- length(x) ny <- length(y) sum1 <- 0 sum3 <- 0 for(i in 1:nx) { sum1 <- sum1 + s...

..._(j-1). In order to estimate y-values, I'm assuming that delta j is approximately equal to kj**u, such that my regression model should be something like this: ^y_1 = a1 ^y_2 = a1 + k2**u ^y_3 = a1 + k2**u + k3**u ... ^y_m = a1 + k2**u + k3**u + ... + km**u or, generically ^yi = a1 + k * sum_j=2^i j**u and I need to fit a non-linear least-squares regression model to find the tripplet a1,k,u. I had a look to the gnm package, but I don't have the lesser idea how to formulate this problem to use this package. Can someone help me with that? cheers, Ronaldo

Dear All, I am evaluating the value of loglikelihood and it ends up with the sum of tiny numbers. Below is an example: suppose I would like to calculate sum_i (log (sum_j x [i, j] )), the index of log (x) is in the range, say (-2000, 0). I am aware that exp(-744.5) will be expressed as 0 in 32 bit R and exp Is there a way to improve the result? R example: powd <- sample(-2000:0, 100, replace=T) # the power of x [i, j] x <- matrix(exp(powd),10)...

...e Ruppert et al textbook on semiparametric regression uses GLS to account for correlated errors. I haven't really used GLS much and I don't think it solves the autocorrelation problem. I'm more accustomed to using a cluster-robust "sandwich" estimator: (X'X)^{-1} (sum_j(X_j' e_j e_j' X_j)) (X'X)^{-1} In a penalized spline context, this would be something like the following: (X'X+\lambda K)^{-1} (sum_j(X_j' e_j e_j' X_j)) (X'X+\lambda K)^{-1} (where J are clusters -- units on whom observations are repeated). As far as I can te...

...j is the outcome (0 or 1) for the j'th case in the i'th cluster It has vector covariates X_ij. c_ij is a transform of y_ij and is in (0, 1) (it's actually the probability of being in cluster j). It may be specified a priori or to be estimated. Maximize the product over i of c_ij exp(sum_j y_ij X_ij b) ---------------------------- sum_K c_ij exp(sum_K y_ij X_ij b) Where sum_K means we are summing over all possible subsets of the strata that contain the observed number of successes. b, and perhaps c, are the parameters to estimate. The more vanilla case has no c_ij terms. That says...

..._(j-1). In order to estimate y-values, I'm assuming that delta j is approximately equal to kj**u, such that my regression model should be something like this: ^y_1 = a1 ^y_2 = a1 + k2**u ^y_3 = a1 + k2**u + k3**u ... ^y_m = a1 + k2**u + k3**u + ... + km**u or, generically ^yi = a1 + k * sum_j=2^i j**u and I need to fit a non-linear least-squares regression model to find the tripplet a1,k,u. I had a look to the gnm package, but I don't have the lesser idea how to formulate this problem to use this package. Can someone help me with that? cheers, Ronaldo

Yes, you're right. In fact, it's just an adaptation of a matlab command and the author advises using N^4 replications that's why it's the default in the function. The bistochastic matrix is not my subject of interest, but I need it to perform some random tranformation of a vector of incomes. Florent Bresson ----- Message d'origine ---- De : Richard M. Heiberger <rmh at

Thanks, I tried someting like this, but computation takes times for large matrices btransf <- function(y,X=length(y)^4) { N <- length(y) bm <- matrix(rep(1/N,N^2),N,N) for(j in 1:X){ coord <- sample(1:N,4,replace=T) d <- runif(1,0,min(bm[coord[1],coord[2]],bm[coord[3],coord[4]]))