Sense about Science (making sense of Forensic Genetics http://senseaboutscience.org/wp-content/uploads/2017/01/Making-Sense-of-Forensic-Genetics.pdf Peter Gill et al) make a number of incorrect statements about defence access to commercial software. I have been in correspondence with Tracey Brown who declined to either make any meaningful changes or substantiate the article’s comments.
In Sense about Science, making sense of Forensic Genetics (http://senseaboutscience.org/wp-content/uploads/2017/01/Making-Sense-of-Forensic-Genetics.pdf), the contributors (Peter Gill et al) state:
“It is important to note that several commonly used computer programs have been developed by commercial companies, and may not be available for use by defence experts because they have less money to pay for them. This could make it difficult to scrutinise how the prosecution generated its statistics. Since open source software is freely available and is more transparent, there is currently a debate within the forensic community on the relative benefits of commercial vs open source software.”
Both TrueAllele and STRmix make their software available to experts acting for the defence in a case free of charge or for a nominal installation charge. STRmix, and we believe TrueAllele have a very defence friendly policy.
20th January 2017. I have heard from a third party that Peter has taken down his presentation. It is a little short of a correction but it is very much appreciated nonetheless. Pete did not inform me personally.
7th January 2017
At writing I have received no response from Peter, a friend of 30 years standing, and usually a fine scientist. In my view this is a disappointing descent below acceptable scientific standards by Peter and I hope it is his last.
Peter Gill presented at the Newton Institute https://drive.google.com/file/d/0B18mPWHZjsgYRkhMOTRGU2JuMTA/view @minute 38 on the case People v Hillary (New York). Gill used material from Mark Perlin’s website which was incorrect. Gill then further corrupted the information by misquoting Perlin. Specifically:
Analytical threshold: Perlin claims that STRmix is validated to 30rfu and suggests that I stopped at 50rfu in the Hillary case for data selection reasons. Gill reverses these numbers and implies in his slide that I went to 30rfu beyond the validated limits. The truth is that neither mention that we have published STRmix papers down to 30rfu on a 3130. The Hillary profiles are run on a 3500 and 50rfu is well low enough.
Different results: Gill states verbally that TrueAllele excludes and STRmix includes. STRmix produced an LR of 300,000. We have no report from TrueAllele and cannot ascertain if we even looked at the same profiles but the Perlin email states “no statistical support” which I would read as inconclusive.
Gill reports Perlin as stating that there are exclusionary peaks between 30 and 50rfu. He omits the word “potentially” which is present in Perlin’s novel on the case. In my view the peaks are one pull-up and one noise.
I have asked Peter to correct these and I have also offered him the profiles. This would allow him to form a more informed view. I await a response.