[Analogue-commits] r166 - in pkg: R inst
noreply at r-forge.r-project.org
noreply at r-forge.r-project.org
Fri Apr 23 16:40:30 CEST 2010
Author: gsimpson
Date: 2010-04-23 16:40:30 +0200 (Fri, 23 Apr 2010)
New Revision: 166
Modified:
pkg/R/predict.wa.R
pkg/inst/ChangeLog
Log:
useN argument/object being created in wrong place meant it was not available for other CV methods than LOO
Modified: pkg/R/predict.wa.R
===================================================================
--- pkg/R/predict.wa.R 2010-04-23 14:37:48 UTC (rev 165)
+++ pkg/R/predict.wa.R 2010-04-23 14:40:30 UTC (rev 166)
@@ -19,6 +19,7 @@
ENV <- object$orig.env
## tolerance options from model
O <- object$options.tol
+ useN2 <- object$options.tol$useN2
## Doing CV?
if(identical(CV, "none")) {
want <- names(object$wa.optima) %in%
@@ -37,7 +38,7 @@
if(identical(CV, "LOO")) {
loo.pred <- matrix(0, ncol = n.train, nrow = n.fossil)
mod.pred <- length(n.train)
- useN2 <- object$options.tol$useN2
+ ##useN2 <- object$options.tol$useN2
want <- names(object$wa.optima) %in% colnames(newdata)
want <- names(object$wa.optima)[want]
nr <- NROW(X) - 1
@@ -218,7 +219,7 @@
retval$model.pred <- list(pred = fitted(object))
} else if(identical(CV, "LOO")) {
mod.r.squared <- cor(mod.pred, ENV)^2
- mod.resid <- mod.pred - ENV
+ mod.resid <- ENV - mod.pred ##mod.pred - ENV
mod.avg.bias <- mean(mod.resid)
mod.max.bias <- maxBias(mod.resid, ENV)
mod.rmsep <- sqrt(mean(mod.resid^2))
@@ -231,9 +232,10 @@
resid = mod.resid)
} else {
mod.pred <- rowMeans(oob.pred, na.rm = TRUE)
- mod.resid <- mod.pred - ENV
+ mod.resid <- ENV - mod.pred ##mod.pred - ENV
s1 <- apply(oob.pred, 1, sd, na.rm = TRUE)
- s2 <- sqrt(rowMeans((oob.pred - ENV)^2, na.rm = TRUE))
+ ##s2 <- sqrt(rowMeans((oob.pred - ENV)^2, na.rm = TRUE))
+ s2 <- sqrt(rowMeans((ENV - oob.pred)^2, na.rm = TRUE))
mod.s1 <- sqrt(mean(s1^2))
mod.s2 <- sqrt(mean(mod.resid^2))
samp.rmsep <- sqrt(s1^2 + mod.s2^2)
Modified: pkg/inst/ChangeLog
===================================================================
--- pkg/inst/ChangeLog 2010-04-23 14:37:48 UTC (rev 165)
+++ pkg/inst/ChangeLog 2010-04-23 14:40:30 UTC (rev 166)
@@ -15,6 +15,8 @@
* summary.predict.mat: print method was incorrectly extracting
the model estimates for training set samples.
+ * predict.wa: fix minor bug with CV when tolerance DW was used.
+
* Package: reduced package check time in examples, by using
the Imbrie & Kipp data.
More information about the Analogue-commits
mailing list