<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta http-equiv="content-type" content="text/html; charset=ISO-8859-1">
</head>
<body bgcolor="#ffffff" text="#000000">
A major update of <b>travelr </b>has been posted to the pre-alpha SCM
section, with some new documentation, working ("at least it runs
without crashing") traffic assignment including all the base algorithms
(notably Frank-Wolfe and ParTan), revised iterative proportional
fitting function (I never could figure out the purpose of R's "sweep"
function -- thanks ODOT'ers! Now ipf can do as many dimensions on your
array as you like...), plus there is a pair of neat (at least, I think
they're neat) new functions:<br>
<ol>
<li>an adaptive quadrature function for computing numerical integrals
of cost functions, which you need to do equilibrium optimization
objective functions -- it's always faster to code the integral
analytically, but the numerical integration adds to the toolbox</li>
<li>a function for re-aggregating matrices using an equivalency
table. So you can convert your census block data to TAZs, or add up
the TAZs into districts, or sub-divide counties into TAZs, etc. This
one is still pretty primitive and may not work smoothly with some
obvious things, such as applying it to two-dimensional margins of
three-dimensional arrays in order to map each sub-table into the same
district structure<br>
</li>
</ol>
There are some new tests and examples (many more to come). Feel free
to try something out and contribute the code back as a new or revised
test: even if (and ESPECIALLY if) the function crashes doing something
obvious, or you can't get it to do what you want, send me code that
reproduces the problem, I'll fix it (either the function or the
documentation), and I'll incorporate the test in the standard
sequence. I haven't formalized enough tests yet, since I'm impatient
to get to the "feature complete" stage (it's very close), but more and
more are suggesting themselves and making their way into the "to-do"
list. I do intend to ramp up R-Forge's "tracker" feature so users can
easily make bug reports, feature requests, request support (not likely
to be highly responsive), etc.<br>
<br>
All my TravelR energy now is going into a couple of large extended
example that will become a vignette (one a whole model, the other a
tutorial on how to code up a new assignment algorithm), plus
incrementally adding to and polishing the documentation and examples
(plus the comments in the code), some of which are a bit thin. On the
other hand, each time I've started digging into that, I've had to
overhaul the code in some way or another. Stability is still a way
off...<br>
<br>
Happy modeling!<br>
Jeremy "R"<br>
<br>
</body>
</html>