-
Can someone explain more about what this means and why this might happen? I'm getting this error in a "leave_sdlg" kernel where seedlings (discrete) enter a mature plant cohort that is structured by size (continuous). It only happens with certain runs of a stochastic IPM as well, so I assume it has something to do with this kernel generating plants that are too small. I thought I was correcting for that with |
Beta Was this translation helpful? Give feedback.
Replies: 3 comments 1 reply
-
|
I think I answered my own question. I had accidentally forgot to back-transform a survival model, so survival was often > 1. Not entirely sure how that relates to the error message, but fixing that seems to have gotten rid of it. |
Beta Was this translation helpful? Give feedback.
-
|
So the error message means that some value in the sub-kernel named leave_sdlg is negative, which can't be true for an IPM to work (i.e. it will produce negative individuals when multiplied by the population trait distribution). If that sub-kernel had stochastic parameters whose distributions could produce negative numbers, or forgetting to back transform yielded negative values on the logit scale, then that would definitely be the cause. |
Beta Was this translation helpful? Give feedback.
-
|
FWIW, it took a long time for the error message to "click" for both me and my advisor. I think we both initially thought the floating point error WAS the error, but it's really saying "floating point errors are cool, but this is not just a floating point error". The way it's currently worded, I think it could be interpreted as saying that there are negative numbers greater than expected and that they were produced by floating point errors. |
Beta Was this translation helpful? Give feedback.
-
|
Yea, I can see how those are not the most user-friendly error messages.
Tragically, I just submitted an update to CRAN last week, so these will
need to get rolled into the next iteration of ipmr. Sorry about that!
…On Fri, May 13, 2022 at 10:50 AM Eric R Scott ***@***.***> wrote:
FWIW, it took a long time for the error message to "click" for both me and
my advisor. I think we both initially thought the floating point error WAS
the error, but it's really saying "floating point errors are cool, but this
is not just a floating point error". The way it's currently worded, I think
it could be interpreted as saying that there are negative numbers greater
than expected and that they were produced by floating point errors.
—
Reply to this email directly, view it on GitHub
<#60 (reply in thread)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AGAC2UJ323ZVZD6EOGRICDDVJZTZXANCNFSM5VYJ5IDQ>
.
You are receiving this because you commented.Message ID:
***@***.***>
|
Beta Was this translation helpful? Give feedback.
So the error message means that some value in the sub-kernel named leave_sdlg is negative, which can't be true for an IPM to work (i.e. it will produce negative individuals when multiplied by the population trait distribution). If that sub-kernel had stochastic parameters whose distributions could produce negative numbers, or forgetting to back transform yielded negative values on the logit scale, then that would definitely be the cause.