I did see those qualifiers, they just didn't make sense. With audio of this quality - what quality - is it good or bad ?
In the Orlando Sentinel article, it states "Owen says the audio from Zimmerman's call is much better quality than the 911 call in the Davalloo case."
The implication being that normally, a match of at least 60% with a normal-quality recording is required to establish identity with reasonable confidence. However, the Zimmerman recordings are at a high enough quality that if the samples tested are indeed from the same source, they should show much better match (at least 90%), so that a match of only 60% using the higher-quality samples would have indicated that they weren't as likely to be from the same source, and the actual 48% match identified in this case means that they're not at all likely to be from the same source.
So the lower the audio quality the more accurate the test becomes? What is this, audio homeopathy?
No, it seems to be a case of the lower the audio quality, the less
confident the results will be. That doesn't mean that recordings can't be analyzed, just that a match becomes that much more obvious with a higher quality recording. [EDIT: In other words, it's easy to say that two muffled sounds resemble each other such that they could have come from the same source. But in order to make the same claim about two clear-as-a-bell sounds, they'd have to sound a
lot more alike than two muffled sounds.]
I'd imagine it's an issue of the number of actual matching datapoints in a pool of a given size. For instance, if you have two pools of ten datapoints each, and six match, you can state with some degree of confidence that the two pools are somewhat alike. However, if you have two pools of 100 datapoints each, and
ninety match, those two pools are a
lot more likely to actually be alike.
Got it. Thanks.
