Xvalid

[Jeffrey Yarus]

I have been using the Xvalid code for months with no issues. However, I am not getting errors and as far as I can tell, nothing on my end has changed.... Here is an example of apiece of the code I have used for months:

Code: Select all

db_noout.db2 <- db_noout.db
Bootstrap <-
  xvalid(db_noout.db2,
         data.model.3dir,
         flag.xvalid = 1,
         neigh.unique)
#Bootstrap


Code: Select all

hist(
  db.extract(
    Bootstrap, "Xvalid.P_PHI.stderr"), 
  nclass = 30,
  main = "Histogram of Standardized Error",
  xlab = "Error Variance in units of Standard Deviation",
  col = "red"
)


Code: Select all

db_noout.db2 <- db.locerase(db_noout.db2, "z")
db_noout.db2 <-
  db.locate(db_noout.db2, "8", "z")
  xvalid(db_noout.db2, data.model.3dir, neigh.unique, flag.xvalid = 2)

db_noout.db2 <- db.locerase(db_noout.db2, "z")

db_noout.db2 <-
  db.locate(db_noout.db2, "8", "z")
db_noout.db2 <-
  xvalid(db_noout.db2, data.model.3dir, neigh.unique, flag.xvalid = 1)

db_noout.db3 <- db_noout.db2
#db_noout.db3


The error message is:

Error in xvalid(db_noout.db2, data.model.3dir, flag.xvalid = 2, neigh.unique) :
unused argument (flag.xvalid = 2)

One thing notice is that I can't find the argument, flag.xvalid in the documentation anymore? Is it gone, do I need to do something different?
Any thoughts on this? I cannot figure out why all of a sudden this has stopped working.

Like I said, this has been working for a long time and now it does not.


Jeffrey

[Jeffrey Yarus]

So it seems that the version change to 14.0.3 also changed the Xvalid functions/arguments... Is there an explanation of how we need to modify code to accommodate these changes? I've included in the original post the portion of my code using the old Xvalid functions/arguments.... I'm hoping you folks can recommend the changes I need to make... thanks in advance...

Jeffrey

[Didier Renard]

Hi Jeffrey
Sorry for such a delay in answering to your question. Quite busy these days.

As you mentioned, the arguments of the function xvalid() have changed in order to offer more flexibility.
In a previous version, all the options were contained in the flag.xvalid whose, which made it difficult to grasp without
reading the help manual.
We have decided to make the options more explicit... This is the reason why we now use 3 flags in order to improve
the legibility of the options:
flag.est: it regards the estimation. It can be either:
0 if you do not wish t store anything
1 if you wish to store the re-estimation error (Z*-Z) (traditional output)
-1 if you wish to store Z* alone ... like for a standard Kriging
flag.std: it regards the variance of estimation. It can be either:
0 if you do not wish to store anything
1 if you wish to store the standardized error (Z*-Z)/S* (traditional output)
2 if you wish to store S* (standard deviation of cross-validation error) ... like in a standard Kriging
flag.varz: it regards the variance of the estimator. This is a new option.

Hope this answers to you question
Cheers