R help - May 2010 - retrieve path analysis coefficients (package agricolae)

Dear list,
I'd like to use path.analysis in the package agricolae in batch format on
many files, retrieving the path coefficients for each run and appending them
to a table. I don't see any posts in the help files about this package or
the path.analysis package. I've tried creating an object out of the call to
path.analysis, but no matter what I try, the function automatically prints
the result. I'll be grateful for any assistance.

Thanks in advance,
Zack

	[[alternative HTML version deleted]]

On 2010-05-23 18:16, Zack Holden wrote:
> Dear list,
> I'd like to use path.analysis in the package agricolae in batch format
on
> many files, retrieving the path coefficients for each run and appending
them
> to a table. I don't see any posts in the help files about this package
or
> the path.analysis package. I've tried creating an object out of the
call to
> path.analysis, but no matter what I try, the function automatically prints
> the result. I'll be grateful for any assistance.
>
The beauty of R is that it is open source software giving
you access to the code.

The function agricolae::path.analysis is very simple.
Unfortunately, it doesn't return anything and just prints
its results. But that's easy to fix. Here's a modified
version that just removes all the printing stuff and instead
returns the values of the path coefficients which you can
extract with '$Coeff' (note the capital 'A' in path.Analysis):

path.Analysis <-
function (corr.x, corr.y)
{
     if (ncol(corr.y) > 1)
         corr.y <- t(corr.y)
     Direct <- solve(corr.x, corr.y)
     n <- ncol(corr.x)
     Coeff <- corr.x
     for (i in 1:n) {
         for (j in 1:n) {
             Coeff[i, j] <- Direct[j] * corr.x[i, j]
         }
     }
     invisible(list('Coeff'=Coeff))
}

The loops aren't really necessary, so here's a simpler version:

path.Analysis <-
function (corr.x, corr.y)
{
     if (ncol(corr.y) > 1)
         corr.y <- t(corr.y)
     Direct <- solve(corr.x, corr.y)
     n <- ncol(corr.x)
     Coeff <- t(corr.x * rep(Direct, n))
     invisible(list("Coeff" = Coeff))
}

   -Peter Ehlers
> Thanks in advance,
> Zack
>
> 	[[alternative HTML version deleted]]
>
> ______________________________________________