R-beta: spline problems(?)

Douglas Bates bates@stat.wisc.edu
07 Apr 1998 13:57:55 -0500


"A. Tsiolakidis"  <tsioltas@cs.keele.ac.uk> writes:

> OK, thanks, it looks like I can use the splines library instead (I had no
> idea how, but now I do!). Should we call it a bug then?
> 
> Question No. 2:
> 
> >From the same S-plus program, I was forced to change a line that was giving
> me Segmentation Error problems (a crash on Solaris 2.5):
> 
>  error _ t(y.fit) %*% ( DBD + Lambda ) %*% y.fit +
>  rbind(rep(1,n)) %*% F.vec.sq - 2 * t(y.fit) %*% F.vec

> Apparently, leaving out the middle term did the trick:
>  error _ t(y.fit) %*% ( DBD + Lambda ) %*% y.fit - 2 * t(y.fit) %*% F.vec
> , while the program still runs more or less as expected. There's nothing
> nasty about this middle term, is there?

It could possibly be a problem with rbind or with the matrix
multiplication for a column vector like that.  It is unusual to create 
a matrix that way.  If you want a single column you could create it as
 matrix( 1, nrow = n, ncol = 1)
or reverse the value of the nrow and ncol arguments for a matrix with
1 row.  Even more concise is
 array( 1, c(n, 1) ) or array( 1, c(1, n)) 
depending on which orientation you want.

> Finally, a certain READ.MAT function, that reads in a single-column file,
> taking the first two entries as the dimension of a matrix and storing the
> remaining values in such a matrix, wouldn't work; anybody can see why?

> READ.MAT _function(file, nrow = 0) {
> if(nrow > 0) {
>     n <- scan(file, n = 2)[2] * nrow
>     tmp <- scan(file, n = n + 2)
>   } 
> 
> ## first point of action: simply read in the file in vector 'tmp'
> else {
>     tmp <- scan(file)
>   }
> 
> if(length(tmp)<=2) {
>     return(NULL) 
>   }
> 
> ## the following clause should do the vector>>matrix conversion
> 
>  else {
>     x <- matrix(tmp[ - c(1, 2)], ncol = tmp[2], byrow = T)
>     return(x)
>   }
> }

I'm not sure exactly what is happening with that.

> Finally, if anybody is curious to see what is giving me all this trouble,
> it is some code for what is termed Mixture Spline models:
> http://cm.bell-labs.com/cm/ms/departments/sia/dxsun/speech/mixSplineSpec

This was sent to me only.  I am forwarding the reply onto the r-devel
list so others can take a look at it.
-- 
Douglas Bates                            bates@stat.wisc.edu
Statistics Department                    608/262-2598
University of Wisconsin - Madison        http://www.stat.wisc.edu/~bates/
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-devel-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._