Skip to content

Latest commit

 

History

History
21 lines (13 loc) · 1.23 KB

README.md

File metadata and controls

21 lines (13 loc) · 1.23 KB

CoCoA: Conditional Correlation Models with Association Size

This R package implements the conditional correlation model in the paper, "CoCoA: Conditional Correlation Models with Association Size" by Tu et al, now in print at Biostatistics. It can be installed via

devtools::install_github("danni-tu/rCoCoA")

The preprint for this paper is also available on bioRxiv.

Conditional Correlation Models

In the CoCoA model, we consider the coupling between two outcomes $X$ and $Y$ as a function of covariates $Z$ and $T$. In particular, we assume a bivariate distribution for $(X,Y)$ given $Z, T$ with the conditional correlation parameter

$$\mathrm{Corr}(X,Y|Z,T) = g^{-1}(\alpha + \beta Z + \gamma T),$$

where $\alpha, \beta$, and $\gamma$ are parameters to be estimated and $g^{-1}(\cdot)$ is a link function which ensures that correlations are between -1 and 1. The rCoCoA package implements three estimators of the the conditional correlation parameters:

  1. Maximum likelihood (get_params_mle)
  2. Restricted maximum likelihood (get_params_reml)
  3. Second-order generalized estimating equations (get_params_gee).