I work on statistical models for the estimation of Darwinian fitness of the partridge pea plant.

The problem

Roughly speaking, Darwinian fitness is a measure of how well an organism passes it’s genetic information on to the next generation. One way to assess Darwinian fitness for the partridge pea species is to count the number of seeds a plant produces. Modeling seed production can be tricky because so many individuals die off before having the chance to produce any seeds at all, so seed count is a heavily skewed distribution.

The model

Aster models are a type of statistical model that not only solves the skewed distribution issue, but allows us to incorporate genetic information into the equation. This is done through random effects. Each individual in the population gets its own random effect to asses the effect of its father and mother on the number of offspring it can produce.

The estimation

When a model includes random effects at the individual level, there are more random effects in the model than there are datapoints. Markov Chain Monte Carlo, paired with a bayesian analysis can help solve some of the complications that come from having more parameters than data.

The report

My MS Thesis