DNA Identification for Scientists: Methods
Scientists examine all the data to assess hypotheses. Modern computers explain how quantitative DNA data arises from underlying biological processes. Mathematical models account for observed data and their random variation. More accurate data models can better explain DNA evidence, and thereby preserve more DNA identification information.
The polymerase chain reaction (PCR) underlying DNA identification is a random process. PCR experiments sample peak patterns from a well-understood data probability distribution. This lecture shows how quantitative modeling exploits data variation to preserve DNA identification information, while qualitative thresholds introduce error and discard information.
Quantitative Data Modeling
A quantitative data model includes many explanatory variables, including genotypes, mixture weight and data variance. These variables (and their uncertainty) are determined by a joint probability distribution over all the data and parameters. This lecture shows how using a quantitative likelihood function can preserve DNA identification information.
Mixture Weight and Inference
Statistical inference of genotypes and other DNA variables follows human intuition about the underlying biological processes, but can apply far more computing power. In this lecture, we use mixture weight as an illustrative example of how computers can refine human insight. We also show how generally accepted computer sampling can account for all the DNA variables.