This repository contains two scripts related to the study of molecular markers of antimalarial drug resistance
- Beta_binomial_percolateR is iterations of a function which take prior genotyping data in the form of prevalence and current genotyping data generated via recent benchwork
Prior information containing count data of both mutant infections, and sample size, were coded via a beta distribution.
Π ~ Beta(α,β) α = Prior mixed + Prior mutant infections β = Prior WT infections
Our observed outcome (prevalence) is distributed binomially, generating the following likelihood function.
Y ~ Binomial(Observed mutant infections, Π)
The posterior distribution of prevalence given our observed data is described via a beta-binomial conjugate prior. P(Π|Y) = Beta(α^1,β^1) α^1 = Prior mutant infection + observed mutant infections β^1 = Prior WT infections + N observed - observed mutant infections
All years which had no corresponding information are parameterized via Jeffreys prior, an uninformative prior of the beta distribution. α uninformative = .5, β uninformative = .5
- k13_TMLE is an implementation of a "causal inference estimator" targeted maximum likelihood estimation for the DAG below, done in order to understand the unidentified impact of indoor residual spraying on the emergence of k13 mutations within Ugandan Plasmodium falciparum populations. This implementation respects district level clustering. This code is annotated for reproducibility, and has been used to implement TMLE in other studies.
Our wished for causal parameter is the mean difference in probability of contracting a P.f. infection with a WHO validated/candidate k13 polymorphism if, contrary to the fact, every study subject lived in district received IRS, vs if, contrary to the fact, no study subject lived in district received IRS.
Ψ*(P*) = E*(Y1) - E*(Y0)