Dear List,
Very many thanks to Nicky Best for the following information in relation to
overwriting temporary data within a winBUGS program.
Martin
*********************************************
You could try writing a 'hard wired' logical function to carry out the
calculations needed to get yconv. It is pretty straightforward to add your
own logical functions to the BUGS language using the WBDev tools (see
http://www1.imperial.ac.uk/medicine/about/divisions/pcphs/eph/bugs/winbu
gs_2.html). This should not only speed up the compilation/run time but will
make your code a lot less messy as you won't need to create lots of
intermediate variables.
Nicky Best
*********************************************
Original Message
I am an occasional winBUGS user, and I do not know the syntax well. I hope
the following question is not too silly. I am working on a convolution
problem, and need to evaluate expressions of the form:
for(index in 1:NROWS)
{
ysum=0.0
for(shift in 1:index)
{
coef = (index-shift) * x1
ysum = ysum + yin[shift] * exp(-coef)
}
yout[index]=ysum*x2
}
But a direct translation of this generic code into BUGS results in multiple
definition errors for coef and ysum. The following code compiles without
error.
for(index in 1:NROWS)
{
for(shift in 1:index)
{
coef[index, shift] <- (index-shift) * x1[jsubj[index]]
ytemp[index, shift] <- ymca[index] * exp(-coef[index, shift])
}
yconv[index] <sum(ytemp[index,1:index]) *x2[jsubj[index]]
}
Congdon (2001, Bayesian Statistical Modelling) uses similar syntax in a
program for regression with autocorrelated errors (Prog. 7.10). But given a
large problem, this would appear to make extravagant use of memory. (Also
takes ages to compile.) Presumably, the extra indexing operations must
result in substantial run-time cpu demands. I can see the reason for
getting the compiler to flag overwrites as an error. Nevertheless, is there
a way of 'overwriting' temporary results, thus replacing the unwanted
matrix/vector ops with scalar ops, and avoiding the resulting overheads?
****************************************************
Dave Lunn makes the following comments on using the WBDev functions coding
approach.
There are three main advantages to doing this: first, function evaluations
associated with 'hard-wired' components can be computed much more quickly
than with their BUGS-language counterparts; second, the full flexibility of
a general-purpose computer language is available for specifying each new
component, and so piecewise functions, for example, can be specified
straightforwardly whereas their specification via the BUGS language (using
the "step(.)" function) can be somewhat awkward; finally, the practice of
hiding the details of complex entities within 'hard-wired' components can
lead to vastly simplified WinBUGS code for the required statistical model,
which reduces the likelihood of coding errors occurring and is easier both
to read and to modify.
Thanks again to Nicky for suggesting this approach,
Martin
*****************
Martin King
Biophysics, Radiology & Physics Unit
Institute of Child Health/Gt Ormond Street Hospital for Children
30 Guilford Street
London WC1N 1EH, UK
tel: + 44 (0)20 7905 2273
fax: + 44 (0)20 7905 2358
email: [log in to unmask]
-------------------------------------------------------------------
This list is for discussion of modelling issues and the BUGS software.
For help with crashes and error messages, first mail [log in to unmask]
To mail the BUGS list, mail to [log in to unmask]
Before mailing, please check the archive at www.jiscmail.ac.uk/lists/bugs.html
Please do not mail attachments to the list.
To leave the BUGS list, send LEAVE BUGS to [log in to unmask]
If this fails, mail [log in to unmask], NOT the whole list
