mixgb - Github version v1.5.3

mixgb 1.5.3

Bug fix

• Fix an error in cbind2(Mis.m, Obs.m) that occurred when the imputed dataset had only a single incomplete variable.
  • Cause of error: In this scenario, Mis.m (a matrix of all other incomplete variables except the currently imputed one) was a 0x0 matrix, which caused error during binding.

- R
Published by agnesdeng about 1 year ago

mixgb - CRAN Version 1.5.2

For CRAN Submission

• Significantly faster imputation by optimising data preprocessing and the use of RcppArmadillo
• Visual diagnostic functions have been moved to the vismi package
• Update Documentation

- R
Published by agnesdeng about 1 year ago

mixgb - Github Version 1.4.0

Optimized mixgb() for large datasets:

• Significantly faster imputation by optimizing data preprocessing and the use of RcppArmadillo
• Enhanced memory efficiency with in-place modifications using data.table
• Bootstrapping option removed from mixgb(). Users can still use bootstrap in the archived function mixgb0().
• PMM is now set to NULL by default.

- R
Published by agnesdeng over 2 years ago

mixgb - GitHub version 1.3.2

Note: We chose to directly release v1.3.2, skipping the release of v1.3.1. The primary changes for both v1.3.2 and v1.3.1 are detailed below. Please be aware that starting from mixgb 1.3.1, the package requires XGBoost version 2.0.0 or higher.

mixgb 1.3.2

Miscellaneous

• Improves package documentation regarding the import of xgb.save() and xgb.load() from XGBoost.

mixgb 1.3.1

Compatibility

• Makes the package compatible with XGBoost 2.0.0 with GPU support:
  • Introduces a new parameter device.
  • Deprecates parameters gpu_id and predictor.
  • Sets tree_method = "hist" by default, aligning with XGBoost 2.0.0.

New Features

• Introduces support for saving imputation models to a local directory through the parameter save.models.folder in mixgb().
  • Models save in JSON format using xgb.save(), a method recommended by XGBoost for future compatibility.
  • When save.models.folder is specified, the return object of mixgb() includes the current imputed datasets, directories for imputation models, and relevant parameters. This object can save using saveRDS() as it doesn't directly contain the models. Users can later load this object into R and employ impute_new(object, newdata, ...) for new data imputation.

- R
Published by agnesdeng over 2 years ago

mixgb - Github version 1.3.1

## Now compatible with XGBoost 2.0.0 - Added new parameter device - Removed depreciated parameters gpu_id and predictor - Set tree_method = "hist" by default, match with XGBoost 2.0.0

## Added feature - Support saving imputation models in a local directory. Users can specify the directory in the parameter save.models.folder in the main function mixgb(). - Models will be save as JSON format using xgb.save() internally. Saving XGBoost models in this way instead of using saveRDS in R is recommended by XGBoost. This can ensure that the imputation models can still be used in later release of XGBoost. - If users specify the save.models.folder, the return object of the main function mixgb() will include the current m sets of imputed datasets, a list of directories of the imputation models, and relevant parameters. Users can save this object using saveRDS() without any concern, as it does not contain the models themselves. For later use, one can load the object into R and impute new data using impute_new(object,newdata,...)

- R
Published by agnesdeng over 2 years ago

mixgb - GitHub version 1.1.0

## Added feature: - Added an auxiliary function default_params() for the main function mixgb(). This function is designed to validate the list of XGBoost hyperparameters that are provided by the user. If users only want to modify a subset of hyperparameter, they can simply pass in those specific hyperparameters as a list in the xgb_params argument, without having to listing all other hyperparameters along with their default values.

## Bug fixes: - Bugs related to PMM for multiclass variables. - Updates plot_hist() and plot_bar() because the dot-dot notation (..prop..) was deprecated in ggplot2 3.4.0. - use after_stat(density) instead of ..density.. in the function plot_hist() - use after_stat(prop) instead of ..prop.. in the function plot_bar()

- R
Published by agnesdeng about 3 years ago

mixgb - GitHub version 1.0.3

• Minor change:
  • Added an auxiliary function default_params() for the main function mixgb(). This function is designed to validate the list of XGBoost hyperparameters that are provided by the user. If users only want to modify a subset of hyperparameter, they can simply pass in those specific hyperparameters as a list in the xgb_params argument, without having to listing all other hyperparameters along with their default values.

- R
Published by agnesdeng about 3 years ago

mixgb - CRAN version 1.0.2

## New version on CRAN * Major change of default settings for mixgb(). - Our package has changed from using bootstrapping to subsampling with a default setting of subsample = 0.7. After further investigations, we discovered that although bootstrapping often works effectively, it can introduce bias in certain situations. As a result, we have made subsampling the default method instead of bootstrapping.

• Current default settings for mixgb().

  • Subsampling (subsample = 0.7)
  • No bootstrapping (bootstrap = FALSE)
  • Minor change:
  • All examples use nthread = 2 to meet the requirement of the CRAN policy.

- R
Published by agnesdeng about 3 years ago

mixgb - Github version 1.0.1

Notable changes :

• use subsampling instead of bootstrapping
• Default settings for the main function mixgb(): subsample = 0.7 bootstrap = FALSE

- R
Published by agnesdeng about 3 years ago

mixgb - CRAN version 0.1.0

Main feature

• First release on CRAN
• Support single and multiple imputation
• Customisable settings regarding bootstrapping and predictive matching
• Visual diagnostics for multiply imputed data

Notes

• Mac OSX users may not be able to use mixgb with multithreading. The package xgboost requires OpenMP to use multi-core. For more information please check https://mac.r-project.org/openmp/.

- R
Published by agnesdeng almost 4 years ago