courtiol / isorix Goto Github PK

On the pdf built by CRAN we have few text overlap with the margin. We need to check that and correct it using hard returns (\cr).

I looked at how one could make nice 3D geospheres in R using rgl.
The following site does not tell the details but the results are stunning, so I really want to add this possibility in the package at some point :-)

link1
link2

It is not that difficult once one has the png of the entire map, I am struggling however to extract the central part of the lattice plots (i.e. the map itself, without the axes, keys, margins...). I want the area selected by trellis.focus()...

Once this will be solved, then the following works:

library(rgl)
map <- rgl.spheres(0, texture="map.png", lit=FALSE)
## for gif movie
movie3d(spin3d(axis = c(0, 1, 0), rpm=2), duration = 30, dir = getwd())

https://www.rdocumentation.org/packages/IsoriX/versions/0.5

Add IAEA info

The month.min and month.max argument do not make much sense at the light of the new approach to fit multiple pairs of models, as one could want just January and August. It would be easier to have a single argument taking a vector of month:
month = 1:12 ## for all
month = c(1, 8) ## for January and August
Let's not forget to remove the Note of the Rmd.

A function leave-one-out has been created a while ago and should be included to the package.

The new function GetElev should be tested on Windows, both with and without tools and Rtools

RandomFields does not seem to be working on some platform:
https://cran.r-project.org/web/checks/check_results_IsoriX.html

So we should move it to suggest and no longer import its functions

Prepare Keith's dataset and run the workflow on it.
Check details with Stefi.

The name is way too long and students get confused...
Maybe keep both names on server for a while.

To be consistent with other functions we should time the calibration and display it once run is done.

When the path is not given, the function forget the final "/" so the generated path does not work.
A possible fix should be to change: paste(path, filename, sep = "") by paste(path, filename, sep = "/")
But since user may give path with final "/" or not and even use "" we should be this function much more robust!

Let's try to see if we cannot find a crude way to do calibration when replicates are lacking... I got this request a few times already...

I have not been perfectly consistent on the coding style.
I will improve this soon and document the coding style people should follow for this project.

It would be nice to be able to handle the projection inside the package.

include new data as in CourtiolRousset2017

The equation plotted by this method is almost certainly wrong...

We need to create an alternative to fitting the calibration model in IsoriX, as calibration function are often recycled from published work. People can provide intercept, slope and residual variance.

AssignDataAlien and CalibDataAlien need to be stored with the correct datasets simulated by the function create_alien()

We should get rid of this object and just build it internally

Shall we change the name of the functions?
Having IsoriX (the package), Isorix (the function) and isorix (the class) may be confusing...
Also using capitals is prone to misuse.
Any suggestions?

A vignette with the statistical definition of each step should be provided with the package.

The Earth distance method has now been transferred into spaMM (v 1.11.70)!

People often need to use an additional shape file.

People want us to weight data by monthly precipitation amounts.

We need a way to combine the assignments performed on different isotopes but on the same individuals.
Maybe Fisher's method with weights...

For now the term 2*cov(predVar, fixedVar)/beta is not included in the variance of the assignment test.
We should try to include it.

not having GhostScript prevents some checks... let's install it.

Change column header in assignment data and make sure ID with spaces work (also check the calibration data and source data). Right now it does not.

The GetElev function uses download.file() and the default method in this function seems to lead to corrupted files on some computers.

• Implementing a checksum test seems the way to go.
• It seems that constraining mode="wb" solved the issue. So we should add an argument mode or method and make "wb" the default setting in GetElev.
• As some individuals have trouble to access to internet, also add optional display using options(internet.info = 0), which provides nice information about connection to internet and shows proxy trials!

A new version of spaMM will be released soon, we must test the new version of the package with it.

The problem with Fisher's method is it assumes tests to be independent which is not true in the case of group migration.
An alternative may be to compute the harmonic mean p-value.

See: https://cran.r-project.org/web/packages/harmonicmeanp/index.html

A vignette explaining how to export spatial objects should be provided with the package

the names of the functions have changed, we must update vignettes and models

Leaflet is a nice system that allows the user to interact with maps.
http://rstudio.github.io/leaflet/

We can add raster easily so it could be nice to implement!

The package does not detach properly, due to the distance method:

library(IsoriX)
detach(package:IsoriX)
library(IsoriX)
## Error : .onAttach failed in attachNamespace() for 'IsoriX', details:
##  call: f(...)
##  error: Entry “Earth” already in registry.
## Error: package or namespace load failed for ‘IsoriX’

The presence/absence of titles on assignment plots is not self consistent.
For group, there is one, for a single individuals none, for several individuals there is.
We should add an argument for the titles to override such defaults.
We may want to change defaults as well...

This is useful if users fit the isoscape directly on animal data.

all functions are done, we just need to show how they work.

I have put all I wanted in onAttach, but this seems to be bad practice:
"Use .onAttach only for actions that are needed only when the package becomes visible to the user (for example a start-up message) or need to be run after the package environment has been created." for ?.onAttach

Add an extra argument in plot to control the border of the mask that is currently wide and black (visible when oceans are not black)

We have to clarify why not having viridisLite installed on the system sometimes create problems.
We are not using viridisLite but we have recursive dependency to it through rasterVis.

It would be good to move all the help content into the R code using Roxygen.
I quite like the namespace as it is, but I don't know if we can keep it with Roxygen.

https://docs.codecov.io/docs/about-code-coverage

add a column (name: groupID) in assigndata for the group assignments. At the moment this is based on lat/long, or if that is not given then all are taken. But it would be great to easily test different groupings, e.g. per species group that one can define, or animals found at the same time......

For some reason the new vignette shows up in CRAN but not in browseVignettes(package='IsoriX')...
I have not yet looked at why this is the case.

RElevate only performs aggregation. We may want to add the resample method, which only takes the real values the closest to the center of the new cells. What do you think?