Taking into account error elipse info in outlier detection

[Stijn]

Dear Chris,

I am working with Argos data with rather large error elipses. I want to remove outliers with the outlie function. 
From attached figure 4 in your 2020 manuscript i understand that the error bars help differentiate uncertain positions from outliers.

From what I understand the following removes positions of higher than threshold speed :

for (i in 1:length(telemetry_object)) {
  OUT <- outlie( telemetry_object  [[i]], plot = FALSE)
  BAD <- OUT$speed  > speed_threshold
  while (any(BAD)) { 
     telemetry_object  [[i]] <-  telemetry_object  [[i]][!BAD, ] 
    OUT <- outlie( telemetry_object  [[i]], plot = FALSE)  
    BAD <- OUT$speed > speed_threshold 
  }
}
  
When I do this, I notice that in my plot positions similar to the example position in figure 4F in your manuscript would also be removed. Is this the right way to do it? Would it not be better to somehow take into account the error bars and so when these error bars cross the speed threshold the position should be kept?

I notice that the OUT object also has a column called VAR.speed. Should this be taken into account somehow?

Thanks a lot for your time and help!

Best, Stijn

[Stijn]

Dear Chris and team,

I puzzled a little further and got a bit more confused about how outlie() uses error information.
When I use Argos error elipses in outlie, I found out that the minimum speed estimates become much larger and unrealistic compared to when I do not use any error info

It seems to me very strange that speed increases with error information? should it not decrease instead?

I attach an example of a subset of data on one of my foxes which illustrates it. it would be much appreciated if you could help me understand a good way to use a speed filter on this Argos data. 

I attach also a screenshot comparison of OUT, with and without error info included.

Thanks aalot!

Stijn

Op dinsdag 21 april 2026 om 19:57:41 UTC+2 schreef Stijn: