This was made to allow people (like me) to specify lower triangular matrices similar to the domain specific language implemented in nlmixr. Originally I had it included in RxODE
, but thought it may have more general applicability, so I separated it into a new package.
For me, specifying the matrices in this way is easier than specifying them using R’s default matrix. For instance to fully specify a simple 2x2 matrix, in R you specify:
mat <- matrix(c(1, 0.5, 0.5, 1),nrow=2,ncol=2,dimnames=list(c("a", "b"), c("a", "b")))
With lotri
, you simply specify:
library(lotri)
mat <- lotri(a+b ~ c(1,
0.5, 1))
I find it more legible and easier to specify, especially if you have a more complex matrix. For instance with the more complex matrix:
mat <- lotri({
a+b ~ c(1,
0.5, 1)
c ~ 1
d +e ~ c(1,
0.5, 1)
})
To fully specify this in base R you would need to use:
mat <- matrix(c(1, 0.5, 0, 0, 0,
0.5, 1, 0, 0, 0,
0, 0, 1, 0, 0,
0, 0, 0, 1, 0.5,
0, 0, 0, 0.5, 1),
nrow=5, ncol=5,
dimnames= list(c("a", "b", "c", "d", "e"), c("a", "b", "c", "d", "e")))
Of course with the excellent Matrix
package this is a bit easier:
library(Matrix)
mat <- matrix(c(1, 0.5, 0.5, 1),nrow=2,ncol=2,dimnames=list(c("a", "b"), c("a", "b")))
mat <- bdiag(list(mat, matrix(1), mat))
## Convert back to standard matrix
mat <- as.matrix(mat)
##
dimnames(mat) <- list(c("a", "b", "c", "d", "e"), c("a", "b", "c", "d", "e"))
Regardless, I think lotri
is a bit easier to use.
lotri
also allows lists of matrices to be created by conditioning on an id
with the |
syntax.
For example:
mat <- lotri({
a+b ~ c(1,
0.5, 1) | id
c ~ 1 | occ
d +e ~ c(1,
0.5, 1) | id(lower=3, upper=2, omegaIsChol=FALSE)
})
print(mat)
#> $id
#> d e
#> d 1.0 0.5
#> e 0.5 1.0
#>
#> $occ
#> c
#> c 1
#>
#> Properties: lower, upper, omegaIsChol
print(mat$lower)
#> $id
#> d e
#> 3 3
#>
#> $occ
#> NULL
print(mat$upper)
#> $id
#> [1] 2
#>
#> $occ
#> NULL
print(mat$omegaIsChol)
#> $id
#> [1] FALSE
#>
#> $occ
#> NULL
This gives a list of matrix(es) conditioned on the variable after the |
. It also can add properties to each list that can be accessible after the list of matrices is returned, as shown in the above example. To do this, you simply have to enclose the properties after the conditional variable. That is et1 ~ id(lower=3)
.