R help - Sep 2009 - < 0 x 0 matrix >

Hi,

Does anybody know, what is going on here?
> diag(sqrt(1))
      [,1]
[1,]    1
> diag(sqrt(0.3333))
<0 x 0 matrix>
> sqrt(1)
[1] 1
> sqrt(0.3333)
[1] 0.5773214

BR, Markku Karhunen
researcher
University of Helsinki

On 09/04/2009 12:25 PM, Markku Karhunen wrote:
>
> Hi,
>
> Does anybody know, what is going on here?
>
>> diag(sqrt(1))
> [,1]
> [1,] 1
>> diag(sqrt(0.3333))
> <0 x 0 matrix>
>> sqrt(1)
> [1] 1
>> sqrt(0.3333)
> [1] 0.5773214
>
> BR, Markku Karhunen
> researcher
> University of Helsinki
Try this instead;

 > diag( sqrt(.3333), 1, 1 )
           [,1]
[1,] 0.5773214

Romain

-- 
Romain Francois
Professional R Enthusiast
+33(0) 6 28 91 30 30
http://romainfrancois.blog.free.fr
|- http://tr.im/xMdt : update on the ant package
|- http://tr.im/xHLs : R capable version of ant
`- http://tr.im/xHiZ : Tip: get java home from R with rJava

Markku Karhunen wrote:
> Hi,
>
> Does anybody know, what is going on here?
>   
diag( x ) produces a round(x) x round(x) identity matrix when x is 
length 1.   (This is the third case listed on the man page ?diag).  See 
the note there about a safer form if you wanted a matrix with x on the 
diagonal.

Duncan Murdoch
>   
>> diag(sqrt(1))
>>     
>       [,1]
> [1,]    1
>   
>> diag(sqrt(0.3333))
>>     
> <0 x 0 matrix>
>   
>> sqrt(1)
>>     
> [1] 1
>   
>> sqrt(0.3333)
>>     
> [1] 0.5773214
>
> BR, Markku Karhunen
> researcher
> University of Helsinki
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>

On Fri, Sep 4, 2009 at 11:25 AM, Markku
Karhunen<markku.karhunen at helsinki.fi> wrote:
> Hi,
>
> Does anybody know, what is going on here?
>
>> diag(sqrt(1))
>
> ? ? [,1]
> [1,] ? ?1
>>
>> diag(sqrt(0.3333))
>
> <0 x 0 matrix>
>>
>> sqrt(1)
>
> [1] 1
>>
>> sqrt(0.3333)
>
> [1] 0.5773214
>
 Read the help for diag yet?

    'diag' has four distinct usages:
...
        3.  'x' is a scalar (length-one vector) and the only argument
           it a square identity matrix of size given by the scalar.
...

So diag(0.1) becomes diag(0) which is a 0-size matrix. Try diag(2.4)

Barry

it's documented as "unexpected"

?diag
Note

Using diag(x) can have unexpected effects if x is a vector that could be of
length one. Use diag(x, nrow = length(x)) for consistent behaviour.


And the result follows from this part,

    else if (length(x) == 1L && nargs() == 1L) {
        n <- as.integer(x)
        x <- 1
    }


baptiste

2009/9/4 Markku Karhunen <markku.karhunen@helsinki.fi>
> Hi,
>
> Does anybody know, what is going on here?
>
>  diag(sqrt(1))
>>
>     [,1]
> [1,]    1
>
>> diag(sqrt(0.3333))
>>
> <0 x 0 matrix>
>
>> sqrt(1)
>>
> [1] 1
>
>> sqrt(0.3333)
>>
> [1] 0.5773214
>
> BR, Markku Karhunen
> researcher
> University of Helsinki
>
> ______________________________________________
> R-help@r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
> http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>


-- 
_____________________________

Baptiste Auguié

School of Physics
University of Exeter
Stocker Road,
Exeter, Devon,
EX4 4QL, UK

http://newton.ex.ac.uk/research/emag
______________________________

	[[alternative HTML version deleted]]

> On 04-Sep-09 10:45:27, Markku Karhunen wrote:
>> True. Should have read ?diag.
>>
>> However, this provokes a more general question: Is there some way I
>> can declare some scalar and _all its functions_ as matrices?
>>
>> For instance, I would like to
>>
>> A = as.matrix(0.98)
>> B = function(A)
>> C = diag(sqrt(B))
>>
>> so that all scalars are explicitly [1,1] matrices.
>> BR, Markku
>
> Hmmm, it might be a good idea to explain why you want to do this.
> For instance:
>
>   M <- matrix(c(1,2,3,4),nrow=2)
>   c <- matrix(2,nrow=1)
>   c%*%M
>   # Error in c %*% M : non-conformable arguments
>   c*M
>   # Error in c * M : non-conformable arrays
>   c+M
>   # Error in c + M : non-conformable arrays
>
> So what would you want to use the [1,1]-matrix scalars for, that
> cannot be done just using them as numbers?
>
> Ted.
Broadly speaking, I would like to use the same code for multivariate  
and univariate cases. For instance, I use the inverse Wishart  
densities of MCMCpack. If I take diwish(x) of a scalar x, the  
programme crashes, because diwish() by default checks  
ncol(x)==nrow(x). However, I would like to have an inverse gamma  
density.

Best,
Markku