R/NRpoisson.R

In isglobal-brge/CNVassoc: Association analysis of CNVs and imputed SNPs incorporating uncertainty

NRpoisson <- function(y, X, w, beta, variant, 
        tol, max.iter, verbose) {
        if (missing(variant)) 
            variant <- !apply(is.na(beta), 1, 
                any)
        beta <- t(apply(beta, 1, function(x) if (any(is.na(x))) 
            rep(x[1], length(x)) else x))
        J <- NCOL(w)
        n <- NROW(y)
        K <- NROW(beta)
        if (max.iter <= 0) {
            derivatives <- hessianPoisson(beta = beta, 
                y = y, w = w, X = X, variant = variant)  ##
            S <- derivatives$S
            H <- derivatives$H
        }
        iter <- 1
        error <- Inf
        while (error > tol & iter < max.iter) {
            derivatives <- hessianPoisson(beta = beta, 
                y = y, w = w, X = X, variant = variant)  ##
            S <- derivatives$S
            H <- derivatives$H
            beta.old <- beta
            beta.oldv <- matrix2vector(beta.old, 
                variant)
            betav <- as.vector(beta.oldv - qr.solve(H) %*% 
                S)
            beta <- vector2matrix(betav, variant, 
                J)
            error <- max(abs(beta - beta.old))
            colnames(beta) <- paste("clust", seq_len(J), sep = "")
            rownames(beta) <- dimnames(X)[2][[1]]
            if (verbose) {
                if (iter == 1) 
                    cat("---- Newton-Raphson procedure ----\n")
                cat("Iter", iter, "Error", error, 
                    "\n")
                print(beta)
                cat("\n")
            }
            iter <- iter + 1
        }
        if (iter > max.iter) 
            warning("Maximum iterations reached")
        return(list(beta = beta, variant = variant, 
            score = S, hessian = H))
}