[R] Undocumented behaviour of diag when replacing the diagonal of a matrix?
Gerrit Eichner
gerr|t@e|chner @end|ng |rom m@th@un|-g|e@@en@de
Thu Dec 5 12:18:08 CET 2024
Hi, Duncan,
thx for your valuable hint re the two assigment functions! That might be
the reason. Nevertheless, I couldn't figure out how they work together
here. I simply executed the relevant lines of code which I found in
`diag<-`body after assigning the matrix to x (which, of course, cannot
yield the same as diag(A[-1,]) <- 7 in the light of your explanation):
A <- matrix(0, nrow = 5, ncol = 5)
x <- A[-1,]
value <- 7
# From `diag<-`'s body:
dx <- dim(x)
len.i <- min(dx)
len.v <- length(value)
if (len.i) {
i <- seq_len(len.i)
x[cbind(i, i)] <- value
}
x
I fiddled a bit around with expressions, but to no avail.
Could you (or anybody else) maybe still shed some more light on this?
Thx a lot in advance!
Best -- Gerrit
---------------------------------------------------------------------
Dr. Gerrit Eichner Mathematical Institute, Room 215
gerrit.eichner using math.uni-giessen.de Justus-Liebig-University Giessen
Tel: +49-(0)641-99-32104 Arndtstr. 2, 35392 Giessen, Germany
https://www.uni-giessen.de/math/eichner
---------------------------------------------------------------------
Am 04.12.2024 um 15:31 schrieb Duncan Murdoch:
> On 2024-12-04 7:38 a.m., Gerrit Eichner wrote:
>> Dear list,
>>
>> is anyone aware of the following behavious of diag when used to replace
>> diagonals (plural!) of a matrix?
>>
>> Small example: The following is documented and clearly to be expected:
>>
>> A <- matrix(0, nrow = 5, ncol = 5)
>> diag(A) <- 1; A
>>
>>
>> BUT, what about the following? When executing the code of `diag<-` line
>> by line, it throws errors. So why does it work?
>>
>> diag(A[-1, ]) <- 2; A
>>
>> diag(A[-5, -1]) <- 3; A
>>
>> diag(A[-5, -(1:2)]) <- 4; A
>>
>
> Could you show us the log of what you did that generated errors? The
> statement `diag(A[-1, ]) <- 2` is pretty complex; it involves two
> assignment functions (both `diag<-` and `[<-`), so you might have tried
> to execute the wrong thing.
>
> Duncan Murdoch
>
More information about the R-help
mailing list