Dear all,
I hope someone can assist with soliving my problems with the DCC function in
the BOOTRES package.
Initially did my analyses in Dendroclim, but when I tried to repeat the
analyses in dcc I encountered some problems, that I think I should report.
First I encountered that that one variable (April precipitation) that was
determined as significant in Dendroclim's response function analysis was not
significant when using the corresponding dcc function, using exactly the
same data and level of significans.
Moreover the values in the output-files of dcc does not reflect completely
the values from Dendroclim (with all decimals).
While Dendroclim returns exactly the same values each time the same analysis
is run, dcc values seem to change a bit each time. This seems a bit odd, but
I do not know the analysis good enough to say it is wrong.
After more testing I found out that 'April precipitation' is significant
first time I run the dcc analysis after starting R, but in all succeeding
runs, the variable is claimed NS. And this despite the confidence intervals
do not always include the zero value in each run, (which according to my
knowledge is the way the significant variables are determined).
One last question: I would like to know if values from correlation analysis
or response function analysis fits the data best. Is there a way to to
obtain a value for model performance from dcc like e.g. a AIC value?
I use :
R version 2.15.1 (2012-06-22) -- "Roasted Marshmallows", Copyright (C)
2012
The R Foundation for Statistical Computing ISBN 3-900051-07-0, Platform:
i386-pc-mingw32/i386 (32-bit). For Windows XP.
But I also encountered the same problems in R version 2.15.0.
My script:
response <- dcc(kronologiprewhite, list(temp,precip), method =
"response",
start = -6, end = 7, timespan=c(1963,2008), ci = 0.05, boot=TRUE)
I very much hope someone can find the time to answer my questions.
Rasmus Halfdan J?rgensen
--
View this message in context:
http://r.789695.n4.nabble.com/Problem-with-Bootres-dcc-and-Dendroclim-match-tp4638327.html
Sent from the R help mailing list archive at Nabble.com.
