1) What data do Hare et al use?
For this study, Hare et al used the random forest package, the x3p file, and bullet r package. In regards to their reference data base, they used the Hamby study as a starting point and analyzed the data from there.
2) In what ways do the methods used by Hare et al differ from the “traditional” methods of bullet matching?
There were multiple things that were done during this study that are different from the usual procedures followed by forensic criminalists. First and foremost, they scanned the bullets and created digital images to assist in their comparisons. As far as I know at this time (based on our talks with multiple bullet examiners) most criminalists working in firearms and tool marks use a comparison microscope and rely on their skill and experience. In addition, they used points of interest in this study beyond consecutively matching striae.
3) How do Hare et al use clustering to help perform bullet matching tasks?
While clustering was not explicitly mentioned in this video, based on the data that was presented they were using a similar method to organize and analyze their data. They took the scans of the different bullets and compared their traits in order to find similarities in their structure and see if they can be mistaken for one another.
4) Identify one statistics and/or probability concept in the presentation that you have not heard of before. Do a little bit of research (Googling/Wikipedia is ok) and try to describe it to someone who doesn’t know about it. You should also consult this paper to see if there is more detail on your chosen topic than is presented in the webinar. (Hint: Control + F is useful…) You don’t need to read the whole paper.
The LOESS regression is also known as a “locally weighted polynomial regression”. Based on the Wikipedia page on this subject, the each individual point if interest is placed on a plot. A least squares equation helps to judge exactly were the point belongs and how important it is to the data as a whole.
The grid is essentially broken up into columns and the data is plotted in its correlated portions. The points in each individual column have a line set to them. The more columns the more specific the data is. A Then, when testing a regression function, the local polynomial is set to the plot and evaluated.
Source: https://en.wikipedia.org/wiki/Local_regression