The fundamental problems for data mining, statistical analysis, and machine learning are: - whether several distributions are different? - whether random variables are dependent? - how to pick out useful variables/features from a high-dimensional data?
These issues can be tackled by using bd.test, bcov.test, and bcorsis functions in the Ball package, respectively. They enjoy following admirable advantages: - available for most of datasets (e.g., traditional tabular data, brain shape, functional connectome, wind direction and so on) - insensitive to outliers, distribution-free and model-free; - theoretically guaranteed and computationally efficient.
To install the Ball R package from CRAN, just run:
To install the development version from GitHub, run:
Windows user will need to install Rtools first.
Three most importance functions in Ball:
bd.test | bcov.test | bcorsis | |
---|---|---|---|
Feature | Hypothesis test | Hypothesis test | Feature screening |
Type | Test of equal distributions | Test of (joint) independence | SIS and ISIS |
Optional weight | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: |
Parallel programming | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: |
p-value | :heavy_check_mark: | :heavy_check_mark: | :x: |
Limit distribution | Two-sample test only | Independence test only | :x: |
Censored data | :x: | :x: | :heavy_check_mark: |
Interaction screening | :x: | :x: | :heavy_check_mark: |
GWAS optimization | :x: | :x: | :heavy_check_mark: |
Take iris dataset as an example to illustrate how to use bd.test and bcov.test to deal with the fundamental problems mentioned above.
virginica <- iris[iris$Species == "virginica", "Sepal.Length"]
versicolor <- iris[iris$Species == "versicolor", "Sepal.Length"]
bd.test(virginica, versicolor)
In this example, bd.test examines the assumption that Sepal.Length distributions of versicolor and virginica are equal.
If the assumption invalid, the p-value of the bd.test will be under 0.05.
In this example, the result is:
2-sample Ball Divergence Test (Permutation)
data: virginica and versicolor
number of observations = 100, group sizes: 50 50
replicates = 99, weight: constant
bd.constant = 0.11171, p-value = 0.01
alternative hypothesis: distributions of samples are distinct
The R output shows that p-value is under 0.05. Consequently, we can conclude that the Sepal.Length distribution of versicolor and virginica are distinct.
sepal <- iris[, c("Sepal.Width", "Sepal.Length")]
petal <- iris[, c("Petal.Width", "Petal.Length")]
bcov.test(sepal, petal)
In this example, bcov.test investigates whether width or length of petal is associated with width and length of sepal. If the dependency really exists, the p-value of the bcov.test will be under 0.05. In this example, the result is show to be:
Ball Covariance test of independence (Permutation)
data: sepal and petal
number of observations = 150
replicates = 99, weight: constant
bcov.constant = 0.0081472, p-value = 0.01
alternative hypothesis: random variables are dependent
Therefore, the relationship between width and length of sepal and petal exists.
We generate a dataset and demonstrate the usage of bcorsis function as follow.
## simulate a ultra high dimensional dataset:
set.seed(1)
n <- 150
p <- 3000
x <- matrix(rnorm(n * p), nrow = n)
error <- rnorm(n)
y <- 3 * x[, 1] + 5 * (x[, 3])^2 + error
## BCor-SIS procedure:
res <- bcorsis(y = y, x = x)
head(res[["ix"]], n = 5)
In this example, the result is:
# [1] 3 1 1601 20 429
The bcorsis result shows that the first and the third variable are the two most important variables in 3000 explanatory variables which is consistent to the simulation settings.
If you find any bugs, or if you experience any crashes, please report to us. If you have any questions just ask, we won’t bite. Open an issue or send an email to Jin Zhu at zhuj37@mail2.sysu.edu.cn