We try to digitize as much data as possible in as transparent a
manner as possible. Therefore, Rpadrino and PADRINO try to
be as noisy as possible about the underlying models, so that potential
issues don’t slip through unnoticed. This behavior may seem unnerving at
first, but it is only in the interest of maximum transparency with
respect to the underlying IPMs. This document describes in longer form
what these messages mean, and where they originate from.
proto_ipm construction
from ipmr-level functions
These messages appear after a call to
pdb_make_proto_ipm(), and originate from ipmr
functions. Usually, they are innocuous and can be ignored. These
include:
Warning: Assuming that all kernels are implemented with the same 'int_rule'.
This means that, for one reason or another, Rpadrino
couldn’t find the specific integration rule for a specific sub-kernel in
PADRINO. In turn, it assumes that all sub-kernels have the same
integration rule. This is OK because the only integration rule is the
midpoint rule (as of now). We are not aware of any IPMs that use
different integration rules for different sub-kernels within the same
IPM anyway, so this assumption seems reasonable.More will be added here as they appear in later versions of
ipmr.
proto_ipm construction
from Rpadrino-level functions
These messages appear as a result of comments left by digitizers in PADRINO itself. Sometimes, these are issues the user must be aware of, and sometimes, they are comments for the digitization team. Examples of the former include:
Frankenstein IPM. This is used to indicate a
situation where parameter estimates are drawn from multiple sources of
data that the authors did not necessarily collect themselves. For
example, Levin et
al. (2019) used data collected from field sites to estimate
survival, growth, and reproduction of Lonicera maackii, and
then used data from the literature to estimate seed germination,
viability, and establishment probability. We highlight this so that
users who wish to consider, for example, seed vital rates as a function
of climate, are careful to check the original publications and
understand that the latitude/longitude and date ranges listed in PADRINO
may not apply to all vital rates in the model.
Same data as <reference>,
Demographic data from <reference>. These indicate
that the underlying demographic data are taken from another reference,
and used in a possibly modified IPM. Therefore, it would be good to
check both publications to make sure the data meet your criteria for
usage.
There a lot of messages related to latitude/longitude coordinates and their precision. Use caution when using these models in contexts where exact geographic references are required!
'ipm_id' <xyz> has resampled parameters, resetting 'det_stoch' to 'stoch'...
These are models for which all parameter values are not known ahead of
time, because some parameters are random variables drawn from known
distributions. Therefore, they take a bit longer to build and run,
because sub-kernels have to be re-built at every iteration using the new
random parameter draws. It is essential that users get this information
up front, because the default of 50 iterations is certainly insufficient
to understand stochastic behavior. The number of iterations can be
increased with the addl_args argument of
pdb_make_ipm().
ipmr-level functions
More will be added here shortly. In the meantime, email Sam Levin if you encounter issues
during IPM construction with Rpadrino.
Rpadrino-level functions
More will be added here shortly. In the meantime, email Sam Levin if you encounter issues
during IPM construction with Rpadrino.