Creates iteration kernels for IPMs. ipmr does not create
these to iterate models, but they may be useful for further analyses.
format_mega_kernel(ipm, ...)
# Default S3 method
format_mega_kernel(ipm, mega_mat, ...)
# S3 method for class 'age_x_size_ipm'
format_mega_kernel(ipm, name_ps, f_forms, ...)
make_iter_kernel(ipm, ..., name_ps, f_forms)Output from make_ipm.
Other arguments passed to methods.
A vector with symbols, I's, and/or 0s representing the matrix blocks.
They should be specified in ROW MAJOR order! Can also be a character
string specifying the call. Parameter set index syntax is supported. When used,
format_mega_kernel will produce as many mega-matrices as there are
combinations of par_set_indices in the proto_ipm.
The prefix(es) for the kernel name that correspond to survival
and growth/maturation of existing individuals. For the model
K = P_age + F_age, this would be "P". Only applies to
age X size models. The "_age" suffix is appended automatically, so
does not need to be supplied.
The names of the kernels that correspond to production of new
individuals, and possibly, how they are combined. For example, a model that
includes sexual (with an "F" kernel) and asexual reproduction (with a "C" kernel),
this would be "F + C". If data come from multiple sites or years,
then this information is supplied using the index syntax (i.e.
f_forms = "F_yr + C_yr"). Only applies to age X size models. The
"_age" index is appended automatically, so does not need to be
supplied.
A list containing a large matrix or many large matrices (when used with
suffix syntax). The names in the former case will be "mega_matrix"
and in the latter case, "mega_matrix_<par_sets>" with the levels of the
grouping effects substituted in.
ipmr does not generate complete iteration kernels, and uses
sub-kernels to iterate models. However, some further analyses are just easier
to code with a complete iteration kernel. This handles constructing those for
simple and general models of all forms. format_mega_kernel is used
internally by make_iter_kernel for general IPMs. The difference
between these two functions is that make_iter_kernel always returns
a list of objects with c(ipmr_matrix, array, matrix) classes,
whereas format_mega_kernel always returns a list of objects with
c(array, matrix) classes. The former has plot() methods while
the latter does not.
I and 0 represent identity matrices and 0 matrices,
respectively. They can be used to fill in blocks that represent either, without
having to create those separately and append them to the model object. The function
will work out the correct dimensions for both internally, and there is no
restriction on the number that may be used in a given call.
For age_size_ipms, the correct form of mega_mat is generated
internally by creating sub-diagonal matrices for the name_ps kernels,
and a top row using the f_forms. If parameter set indices are part of the
model, the indices should be attached to the name_ps, f_forms in the
function arguments, and the correct block matrices will be generated internally.
data(gen_di_det_ex)
big_k <- make_iter_kernel(gen_di_det_ex,
mega_mat = c(0, go_discrete,
leave_discrete, P))
char_call <- c(0, "go_discrete", "leave_discrete", "P")
big_k_c <- make_iter_kernel(gen_di_det_ex, mega_mat = char_call)
# Now, with an Identity matrix instead of a 0
big_k <- make_iter_kernel(gen_di_det_ex,
mega_mat = c(I, go_discrete,
leave_discrete, P))
# For simple IPMs with no grouping effects, this computes the sum of
# the sub-kernels (i.e. K = P + F)
data(sim_di_det_ex)
simple_k <- make_iter_kernel(sim_di_det_ex)