java.lang.Error: randomChoiceUnnormalized falls through -- negative components in input distribution?

[Sikandar37]

Dear All,
I am running discrete Phylo-geographic analysis for 2600 influenza sequences using fix tree topology following this tutorial https://github.com/rbouckaert/FixedTreeAnalysis. After running the xml file in the beast V.2.7.4/V 2.7.5. i am receiving this error java.lang.Error: randomChoiceUnnormalized falls through -- negative components in input distribution?
Please report error to Marc
What could be the possible reason as previously this tutorial was working well for my data.
expert opinion is needed to solve this error

Thank you

[rbouckaert]

Normally, this happens when a random value is requested from a discrete set of values based on the probabilities of these values. However, when the probabilities are not properly defined (all zero, or all NaN=not a number), the Randomizer cannot pick a value and displays an error.

At first sight, it looks like this happens only once in the whole chain, perhaps due to some numerical instability that makes the probabilities all NaN. Does it ever happen again? Does it seem to make a difference on the posterior? If not, it is probably safe to ignore the message.

[Sikandar37]

Dear Ramco, Thank you for your response, Actually I am getting this error more than once( after every20,000 runs this error is poping up during the whole process. ( Previously I used the same analysis and I achieved the convergence of all the parameter at 5million MCMc and now I am getting this error and this is affecting the the posterior. And taking more time to converge . Is there any way to properly define the probabilities? Thank you for your time

…

On Mon, Dec 4, 2023 at 4:43 Remco Bouckaert ***@***.***> wrote: Normally, this happens when a random value is requested from a discrete set of values based on the probabilities of these values. However, when the probabilities are not properly defined (all zero, or all NaN=not a number), the Randomizer cannot pick a value and displays an error. At first sight, it looks like this happens only once in the whole chain, perhaps due to some numerical instability that makes the probabilities all NaN. Does it ever happen again? Does it seem to make a difference on the posterior? If not, it is probably safe to ignore the message. — Reply to this email directly, view it on GitHub <#1135 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AU7RMGFFRHO6OBSHMJYZHBLYHTI5ZAVCNFSM6AAAAABAD3ADQ2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQMZXGU4DAMRUGI> . You are receiving this because you authored the thread.Message ID: ***@***.***>