Changelog • lukesRlib

lukesRlib 0.2.11 (30th Sept 2024)

tidy_ci() now recognises and tidies factor naming if haven::as_factor() is used instead of the base R as.factor()
get_assoc() now uses haven::as_factor() instead of base R as.factor() for categorical exposures so that labels are shown nicely
Update dependencies to include {haven} and be more explicit which tidyverse packages are desired
Update example data categorical variables to include labels
Fix NEWS formatting

Change URL to reflect my GitHub username change from lukepilling to lcpilling to be more consistent between different logins, websites, and social media
- https://lcpilling.github.io/lukesRlib
- https://github.com/lcpilling/lukesRlib
get_assoc()
- Add options to Winsorize the exposure or outcome
- Remove beep
- Improved verbose reporting

get_assoc() - Fix fit_stat_se if missing
get_assoc() - Fix covariate checking if interaction included in formula. Resolving issue #1

get_assoc()
- add interacts_with argument. Default is ““. User can provide a variable name (string) to also get interaction terms
- Fixed bug in return_all_terms when exposure was factor
- Include SE for C-statistic if returning fit_stat for coxph() models

get_assoc() - add return_all_terms argument. Default is false. If TRUE, adds a new col terms and returns the estimates for all independent variables in the models.
get_assoc() - remove subset_d argument
get_assoc() - fixed if exposure or outcome needs to be protected by backticks (previously failed if variable was something like “bmi-1”)
get_assoc() - fixed problem where covariate estimates were also included in certain situations (e.g., if x=“age” do not get covariate “percentage”)
Update examples for many functions

get_assoc() - fixed if no covariates provided
get_assoc() - if x and y are the same, skip (useful if providing same list of variables twice for pairwise analysis)
get_assoc() - added verbose option for bug checking
get_assoc() - added beep option for beeping when finished
tidy_ci() - add some functionality for summarizing fixed-effects from lme4::lmer()
Add example_data object for examples. Contains ~5,000 participants from Framingham Heart Study

get_assoc() now expects the data as first argument, if not defined. Brings it more into the tidyverse. For instance, one can now:
- data |> filter(age>60) |> get_assoc(x="BMI", y="diabetes", z="sex")
Add theme theme_minimal_modified for ggplots
Moves most dependencies to “Imports” - this is the preferred way to use existing packages https://r-pkgs.org/dependencies-mindset-background.html#sec-dependencies-namespace
Fix -log10p calculation

Add option “get_fit” to get_assoc() to also get model fit statistic when PheWASing. For each model type the fit_stat is:
- lm = R-squared
- logistic = McFadden’s pseudo-R2
- coxph = Harrell’s c-statistic
Fix error if user does not want to provide any covariates to get_assoc()
tidy_ci() now returns the C-statistic for CoxPH models by default

Decided to remove genetics-related functions to instead have them in a separate package, as the dependencies were getting out of hand
- See https://github.com/lukepilling/gwasRtools
Add annotate_textp() function for use in ggplot2 plots for adding easy text annotation geoms
Moved some functions to be internal only to tidy up the namespace
Fix get_p_extreme if NaN is passed rather than NA

Decided to remove doCoxSplinePlot() because it is so much easier to do in {ggplot2} that a whole function is no longer necessary
- Therefore the {pspline} dependency can be removed
Set extreme_ps default to FALSE in get_assoc()

get_assoc() can now perform coxph() models from the {survival} package. The user provides strings depicting the survival object as the outcome e.g., “Surv(time, event)”
Added two genetics-related functions: lambda_gc() (to calculate Lambda GC!) and get_loci() (to crudely determine independent loci in GWAS summary stats)
General improvements/clarifications throughout code and documentation

Added dependency {rlang} >= v1.0.0 so that the sym() function can be used - i.e., using provided variable name strings in dplyr::mutate() etc
get_assoc() now allows options inv_norm_x and inv_norm_y

Changed get_assoc() so that x and y can be vectors - i.e., multiple exposures/outcomes can be provided. Then uses {purrr} function map2() to iterate over combinations
- This package now depends on {purrr} >= v1.0.0 & {stringr} >= v1.5.0
- Added functions xv() and yv() to improve ease of using {purrr} function map2() with get_assoc() for PheWAS of multiple exposures and outcomes in one analysis

fixes to get_assoc() (when no events in exposure group, or when two factor variables included)
add scale options for exposure and outcome to get_assoc()
in tidy_ci() tidy variable names with “scale()”

tidy_ci() now tidies up terms included as categorical variables. “as.factor(x_var)1” is replaced with “x_var-1”. Can be disabled with tidy_factors=FALSE
addition of “Plotting” category of functions, starting with doCoxSplinePlot() that nicely plots the output of a spline term in a CoxPH model
Minor fix to tidy_ci() to avoid breaks due to missing p-values in model summary statistics

Add new function get_assoc() to easily run PheWAS-like analysis using tidy_ci()

Updates to tidy_ci()
Add get_se() from CIs function – very simple but adds to collection of related functions
General improvements

Updated method for -log10 p-value calculation. One just uses Z. Other uses Z and N.

Added two new p-value related functions – very simple but nice to have available