Changelog • NetCoupler

NetCoupler (development version)

CRAN release: 2022-04-08

Added to nc_estimate_* function output the full model list as an attribute, that is really only necessary for those interested in the underlying models used for classifying the effects
Added a continuous outcome variable to simulated data that also links in with the DAG so that the linkage is more obvious (#97)
Added function to create an edge table (#117)
Incorporate tidyselect helpers into functions for selection of variables (#62)
Added Getting Started vignette and an article on examples of using different models (#70)
Added argument to nc_estimate_*_links() functions to set thresholds for classifying links (#157)
Added weights to be included to as_edge_tbl() (#142)

Removed nc_classify_effects() and nc_filter_estimates(), merged them into the two main estimation functions instead
Model summary statistics for lm and glm models were removed for improving computing speed (they slowed things down quite a bit)

Output all models used for classification as an attribute for the nc_estimate_* functions output
Use lavaan instead of dagitty to generate the simulated data
Use standard GitHub Actions and remove AppVeyor
Refactored some code within estimation method so it runs faster
Tidied up the unit tests to run faster
Removed duplicate or extra roxygen examples and instead referenced a common source with @seealso
Removed survival dependency
Switch to using main instead of master branch

For lm and glm models, model summary statistics are added (#88).
Add a function to classify the direct effects between outcome or exposure and the network (#98).
Add function to plot network graph: nc_plot_network() (#89, #110).
Added helper functions nc_adjacency_graph(), nc_adjacency_matrix(), and nc_partial_corr_matrix() to help create the weights for the network plot. (Issue #80, PR #89).
Removed soft deprecated functions. Using MuMIn over glmulti doesn’t change the results too much, see #60 for details (#83).
Removed stringr dependency (#65, #83).

Fix bug where too many digits caused a problem for pcor() (#125, #131).
Fix bug that didn’t properly filter variables nor identify neighbour nodes in nc_filter_estimates() (#109).
Fix problem with nc_standardize() that prevented the ability to use the .regressed_on. argument to extract residuals (#108).
Input dataset can include missingness. Input data is treated as complete case for only the variables used in the modelling (#88).

Rewrote underlying model estimation algorithm so it doesn’t use MuMIn and so there is one unified function for both outcome and exposure side estimation (#101)

Add nc_standardize() function to standardize the metabolic variables (#73).
Export tidyselect functions like matches() or starts_with() (#73).
Add CONTRIBUTING guidelines (#56).
Add lifecycle badges to functions, soft deprecating net_coupler_out(), getExp.coef.permetabolite(), and getExp.coef.out() (#59)
Add defensive checks to input arguments with assertive.types (#59).
Add AppVeyor to repo. Started Travis to run on repo (#61).
Added function for exposure side estimation: nc_exposure_estimates()

Major revision of underlying code for generating the outcome-network link estimation (#55), resulting in created and streamlined nc_outcome_estimates() function. Because of this streamlining, the code is much faster and with the move to use MuMIn we can remove our dependency on rJava via glmulti.
Tidied up nc_create_network() function so that only the graph skeleton is output (#55).
Started cleaning up, along with leftover files.
Updated and generated documentation of nc_create_network().
Added unit tests for nc_create_network() and the outcome estimation functions. Travis and code coverage were added as well.
Renamed nc_make_network() to nc_create_network() and moved into own file.
Modularized nc_make_network() code and moved into another file.