Double peaks in the posterior distribution

[shuyangma1992]

With the newest version of jags and JABBA, I always get strange posterior disributions with two peaks, even I just run the example code. I wonder if this problem is related to the version of these softwares? Or any other reasons?

[Henning-Winker]

Thank you. I am only seeing this now. Good point. I am also seeing this now with example. I will explore. It might have to do with the latest JAGS version but I will also check if this has to do with any of the additions I had pushed recently - although these should not effect the posterior of K or r.

[Henning-Winker]

Fixed version 2.3.5! Although I could not track the origin - updates Windows, R, jags? - something seems to have changed in the numerical space - I found a seemingly robust solution to internally rescale catch and biomass in fit_jabba() as a generic solution, which runs so far very stable on several case studies.

Please check test example
https://github.com/jabbamodel/JABBA/blob/master/Example/Examples_iccat.R

[Henning-Winker]

Big thank you to Rodrigo Sant'Ana, here some further explanation on the bimodal posterior behavior:

This behavior started after the R update last year. I observed this in the last BET evaluation as well. It is probably not a a JAGS or Operating System issue, since I haven't updated JAGS and I use Linux.

Although the background engine is JAGS, which we access via R, it seems to be linked to the machine simulation process, which may have changed in the R update. Or even in the structuring of the distributions. This is why rescaling the captures tended to solve it. Just as increasing the Markov chains also tended to solve it. It seems that at the original scale, the Markov chain requires more effort for singular convergence.

[shuyangma1992]

Thanks a lot for this! 😀

…

________________________________ From: Henning Winker ***@***.***> Sent: Thursday, May 14, 2026 4:47 PM To: jabbamodel/JABBA ***@***.***> Cc: Ma, Shuyang ***@***.***>; Author ***@***.***> Subject: Re: [jabbamodel/JABBA] Double peaks in the posterior distribution (Issue #28) [https://avatars.githubusercontent.com/u/15005753?s=20&v=4]Henning-Winker left a comment (jabbamodel/JABBA#28)<#28 (comment)> Big thank you to Rodrigo Sant'Ana, here some further explanation on the bimodal posterior behavior: This behavior started after the R update last year. I observed this in the last BET evaluation as well. It is probably not a a JAGS or Operating System issue, since I haven't updated JAGS and I use Linux. Although the background engine is JAGS, which we access via R, it seems to be linked to the machine simulation process, which may have changed in the R update. Or even in the structuring of the distributions. This is why rescaling the captures tended to solve it. Just as increasing the Markov chains also tended to solve it. It seems that at the original scale, the Markov chain requires more effort for singular convergence. — Reply to this email directly, view it on GitHub<#28 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/BCSKOJXRM22NVJG74CLR2VL42XMBRAVCNFSM6AAAAACY2NBONOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHM2DINJRG42TINRVGI>. Triage notifications on the go with GitHub Mobile for iOS<https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android<https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>. You are receiving this because you authored the thread.Message ID: ***@***.***>