I am a proud member of the open science revolution. I believe that, within limits (i.e. after publication), we should all share our data and analyses. Sharing data and scripts allows scientists to validate each others research, build off previous work, provides transparency which is often hard to detail in dense scientific papers, and allows for novel use of previously published data. To that end, I will strive to make all my data and analyses available here.
A basic hidden markov model (HMM):Imagine that you really want to know something (e.g. local ancestry across a genome), but you don’t have direct access to that information. What you do have is a series of observations from which you might be able to infer this hidden state (e.g. genotypes at a locus across the genome). This is where you can use a hidden Markov Model.
Here is an R script, in which a professional coin flipper switches between using a fair and a biased coin. I wrote this with the help of CPB graduate students, Alisa Sedghifar, Nick Fabina, Jamie Ashander and Carl Boettiger at our algorithm discussion group. We generate a series of observations of heads and tails to guess which coin he has flipping throughout our encounter. For now we assumed that we knew the rate of switching and the coins bias, but we can do without this prior knowledge by using the Baum Welch algortihm (details later).
With Stephen Wright and Graham Coop, I am currently using an HMM to infer ancestral haplotypes of individuals of the self pollinating species, Capsella rubella across the genome.
Covariance between organelle and symbiont genomes: