README.md

In baruuum/btoolbox: random toolbox

btoolbox

This is just a random collection of functions for coding in R with no specific focus.

Notes:

1. There should be no problems in installing the package with a clang compiler. However, some functions in the Armadillo library rely on OpenMP for multithreading. As clang does not support OpenMP, all Armadillo functions will run sequentially. If you want to enable multithreading via OpenMP, you might follow the instruction on this website.

2. While all functions were tested, some were tested more rigorously than others. The scripts for the unit-tests can be found in ~tests/testthat directory.

Regression

• predict_ci: Calculating confidence intervals for predictions based on GLMs. For polr and multinom objects, Monte Carlo simulations are used; for lm and glm objects, endpoints of the confidence interval of the linear predictor are inverted.

Graphs and Networks

• lr_modularity: calculates the LinkRank modularity (Kim et al. 2010) of a graph partition.
• adj_sum_partition: sums within-partition weights of adjacency matrix
• adj_sum_partition_sp: same as adj_sum_partition but optimized for sparse adjacency matrices
• adj_to_dyadlist: Transforms an adjacency matrix into a dyad-list (not to be confused with a edge-list)
• symmetrize: symmetrizes a matrix by either copying the upper-triangle into the lower-triangle or vice versa
• reciprocated: creates an indicator of whether a tie is reciprocated from an edge list
• update_affilliation: Updates a sparse affiliation matrix (group-by-group) with data on ego-networks.
• sp_eigen_adj: Extracts eigenvalues/vectors from adjacency, normalized adjacency, Laplacian, and normalized Laplacian matrices based on a sparse adjacency matrix.
• components_adj: Extracts the components of a graph based on its sparse adjacency matrix.
• is_connected_adj: Determines whether a graph is connected based on its sparse adjacency matrix.

Clustering and Distances

• mutual_info: Mutual information based on a cross table of frequencies. Available meausures are:
  1. the mutual information between the row and columns of the cross table
  2. Five different forms of normalized mutual informations (NMI)
  3. Five different forms of adjusted mutual information (AMI)
• Distance/similarity measures between each pair of rows of two matrices:
  1. row_pdist: Minkowski distances
  2. frow_euc_dist: fast (but numerically less stable) calculation of Euclidean distances
  3. row_cos_sim: cosine similarity
  4. frow_cos_sim: fast (though numerically less stable) calculation of of cosine similarity

Descriptive Statistics

• sample_logp: Same function as sample() but takes a log-probability vector as argument. The vector might be normalized or only given up to a proportionality constant.
• run_mean: calculates the running average over a numeric vector for different window sizes
• reorder_labels: reorders the labels of an integer, character, or factor variable according to the frequencies with which the labels occur
• quantilize: categorizes a numeric vector into a factor using a number of quantiles as cut-points
• find_interval: finds within which interval a value falls for a variable created by quantilize
• colMedians: calculates the median of the columns of a numeric matrix
• rowMedians: calculates the median of the rows of a numeric matrix
• double_center: double-center a matrix
• gen_agreemat: creates an agreement matrix by comparing the columns (or rows) acrros rows (or columns) and counts the number of entries on which they have the same value

Exported Miscellaneous R Functions

• log_sum_exp: stable calculation of log(sum(exp(x))), where x might be a vector or matrix
• log_accu_exp: stable calculation of log(cumsum(exp(x))), where x might be a vector or matrix
• log_add_exp: stable calculation of log(exp(x) + exp(y)), where x and y are real numbers
• softmax: stable calculation of exp(x) / sum(exp(x))
• log_softmax: stable calculation of log(exp(x) / sum(exp(x)))
• inv_logit: inverse-logit function (logistic transform)

Exported Miscellaneous R Functions for Printing

• round_to_char: rounds a numeric vector to a character vector keeping trailing zeros
• df_to_message: prints data.frame/data.table as a message to the frequencies with which the labels appear
• upper_first_char: changes the first character of a string (or vector of strings) to upper case

References

Kim, Youngdo, Seung-Woo Son, and Hawoong Jeong. 2010. “Finding communities in directed networks,” Physical Review E 81(1)

baruuum/btoolbox documentation built on Aug. 17, 2020, 1:29 a.m.