Bill Northcott has kindly provided a Mac installer for JAGS 1.0.4. This release provides no new functionality but fixes a security vulnerability in the libtool dynamic module loader. A new readme file explains the details.
JAGS 1.0.4 is released. I did not plan this, but I had to fix a potential security vulnerability in libltdl, which the command line interface uses to load modules. Thanks to Dirk Eddelbuettel, who maintains the Debian package for JAGS, for letting me know.
Meanwhile, I am trying to put the finishing touches on the glm module for JAGS 2.0, which should improved mixing in generalized linear (mixed) models. When this is done, JAGS 2.0 should be ready for release.
I will let everyone know when JAGS 2.0 is ready for beta testing.
Bill Northcott has updated the Mac Os X Installer for JAGS. It now works with the version 10.6 (Snow Leopard).
The Mac Os X installer for JAGS 1.0.3 required R to be installed before JAGS. This may be inconvenient for users who do not wish to use the R interface (the rjags package). So Bill Northcott has created an updated installer which does not require R. It is now available on the main page.
If you have already installed JAGS 1.0.3, there is no need to modify your installation
I hope you read the disclaimer that is printed when you launch JAGS - you know the one about ABSOLUTELY NO WARRANTY - because it means what it says.
JAGS 1.0.3, which has just been released, fixes a serious bug in the probit link function, and its inverse, the phi function. These functions returned results based on the upper tail, not the lower tail, of the normal distribution function. Consequently, all the coefficients in probit regression models have the wrong sign. Thanks to Richard Morey for sending me a simple example which showed up the problem.
Obviously, this bug should have been found and fixed well before version 1.0, and I am very embarassed to be revealing it now. Unfortunately, the incorrect results had become entrenched in the benchmarks and so were self-perpetuating. A new release of the classic bugs examples includes updated benchmarks for the BEETLES and LSAT examples.
JAGS 1.0.2 is now out. This is a patched release for the stable version, so there are no major changes in library functionality. The major changes are:
Bill Northcott has kindly provided a binary package of JAGS for Mac Os X 10.4 (Tiger) and 10.5 (Leopard). You can download it from the JAGS home page. Bill writes:
The attached MacOS disk image package contains JAGS, rjags, the manual and the examples. The code is compiled to run on Tiger 10.4 but I have not had a chance to test that as all my computers have been upgraded. It will certainly work on Leopard for both PowerPC and Intel Macs.
The JAGS binary is three way fat (i386, x86_64, and ppc) which means it runs as 64 bit on a Core2Duo Mac. This works perfectly with R 2.6.1 and rjags which are only 32 bit.
Several people have written in to tell me that rjags crashes if you re-run a modelling script in the same R session. The crash is caused by a memory management bug in the JAGS library that is triggered when you overwrite a jags model object with a second call to the jags.model() function.
JAGS-1.0.1 fixes this problem. Since the rjags version is tightly linked to the JAGS release number, you will also need to install rjags-1.0.1, even though this has no code changes compared with rjags-1.0.0.
People using only the command line interface can carry on using JAGS 1.0.0.