3
$\begingroup$

I am working on a TLE filtering code to prepare the TLE series for orbit propagation. I am using a sliding window approach with meadian+ n*mad or IQR to detect the outliers in those orbital elements but this approach requires that for every TLE and also for every element human intervention is required to determine the threshold for the outliers. I would like to make it more autonomous in setting an optimum threshold but I'm not very good with statistics. I used the work by Lidtke as a guide, but I cannot follow very well what he did. Does anyone knows any similar works, or any code examples that I could study on?

$\endgroup$
1
$\begingroup$

What exactly is your goal? Are you looking for errors in the TLE data, or are you trusting them to be correct and using their distribution to tell you what sorts of orbits are unusual for satellites to be in? Trying to spot errors just by statistical means is inherently murky; if you want to do that for real, you need better orbit data to compare against. If you just want to flag things to look at more closely, { X < Q1-1.5 IQR } U { X > Q3+1.5 IQR } is a pretty standard rule of thumb. The only way to decide what threshold would be optimal is to have some other way to tell you which of your categorizations are right or wrong, and find an N that balances in some chosen way between Type I and Type II errors.

$\endgroup$
1
  • $\begingroup$ I'm looking to find anomalies in their distribution as some of them might be erroneous or mark an event. I have already used the IQR approach, but I'm looking for some tunable methods. $\endgroup$ – vasea Sep 25 '20 at 14:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.