Error when checking for outliers: Error in arima(.Lons.ts, order = c(1, 1, 0)) : non-stationary AR part from CSS

[VirginiaMorera]

Hello all,

 

I am testing the code of the FLightR package and I run into an error when checking for outliers.

 

When running

 

Outliers<-detect.tsoutliers(Calibration, Proc.data, plot=T, Threads=Threads, max.outlier.proportion=0.075, simple.version=F)

 

where

 

Calibration is the object I've generated with the create.calibration() function

Proc.data is the output of process.twilights()

and the rest of the parameters are set as in the example workflow in GitHub

 

I get the following:

 

making cluster
estimating dusk errors projection on equator
estimating dawn errors projection on equator
reestimating dusk errors projection on equator
detecting outliers
Error in arima(.Lons.ts, order = c(1, 1, 0)) :
non-stationary AR part from CSS

 

And I've got absolutely no idea of what's going wrong. I've tried with a calibration period of just a few days (7) and also with a calibration period that is the entire breeding season, but both crash at the same point.

 

I'm trying to replicate the "Intigeo tag on a Bar tailed godwit analysis example" workflow with version 0.3.6 in an R version 3.3.1

 

Any help would be much appreciated, thanks!

 

Virginia

[Eldar]

   Hi Virginia,

 It is hard to tell what is going on there without having your data and script, but I would not go to far in search for the error source and use the latest FLightR version (it is currently 0.4.3). To install the latest try 

library(devtools)
install_github('eldarrak/FLightR')

  And here is an updated workflow.

  Hope this helps,

 Eldar