This R package implements the CAUSE method described in Morrison et al 2019 (BioRxiv https://www.biorxiv.org/content/10.1101/682237v4). Get started with an example analysis: https://jean997.github.io/cause/ldl_cad.html
We have discovered that there is an error introduced in cause elpd computations using newer versions of the loo package with cause version 1.1.0 due to a change in the order of the loo_compare
output.
If you are new to CAUSE please use v1.2.0 or the most recent github version. We will issue a new release soon.
If you have used cause version 1.1.0 or used cause with loo version 2.3.0 or 2.3.1 please do the following steps.
- Insall v1.2.0 or the most recent version from github.
- Recompute the elpd model table using the
recompute_elpd_table
function.
elpd_table <- recompute_elpd_table(res)
res$elpd <- elpd_table
Here res
is an object produced by running the cause function. You may find that the new table and the old table are the same or the sign of the z-score comparing models will flip, changing the p-value. The error only occurred in situations when the sharing model is better than the causal model.
- Posterior estimation algorithm is updated which should be faster for traits on very different scales: Version
1.2.0.0331
and later. - The
pval_cols
option ingwas_merge
is available as of version1.2.0.0320
.
- Corrected bug that came from re-ordering of loo_compare output in >= 2.3.0.
- Added function to recompute z-score for previously run analysis.
- Minor updates to parameter checking in gwas_merge.
- Compatible with current versions of mixsqp (0.3-43) and ashr (2.2-47). This version is no longer compatible with versions of mixsqp prior to 0.3.
- Compatible with loo 2.3.0 (no more warning about
compare
being depricated). - elpd object no longer contains comparisons with null model.
- Improved documentation and arguments for cause function. Added the pval_thresh argument.
The original version of the cause
R package is only compatible with earlier versions of mixsqp
and ashr
. The latest version is compatible with newer versions of those packages. If you want to exactly replicate the results in the paper you should use version 1.0.0. It is possible that the newer version is slightly faster. This gives two installation options:
devtools::install_version("mixsqp", version = "0.1-97", repos = "http://cran.us.r-project.org")
devtools::install_version("ashr", version = "2.2-32", repos = "http://cran.us.r-project.org")
devtools::install_github("jean997/cause@v1.0.0")
Don't allow R to update mixsqp
or ashr
.
For the release version
devtools::install_github("jean997/cause@v1.2.0")
For the development version
devtools::install_github("jean997/cause")
Note that for CAUSE versions after 1.0.0, you must use the newer version of mixsqp
(0.3.XX)