Table 1 Description of feature filtering methods

SD-b

Standard deviation based on beta values

SD = sqrt(1/N ∑(Xi-mean(X))²)

SD-m

Standard deviation based on M values

SD = sqrt(1/N ∑(Xi-mean(X))²)

MAD

Median absolute deviation of beta values

median(|Xi-median(X)|)

DIP

Measure of unimodality in a sample

The max difference, over all sample points, between the empirical distribution function and the unimodal distribution function that minimizes that maximum difference

Precision

Inverse precision parameter

1/phi = 1/(mean(X)(1-mean(X))/SD²-1)

BQ-GOF

Beta Quantile Goodness-of-fit

Sum the absolute differences, over 25 quantile points, between the empirical distribution function and the expected beta distribution function

TM-GOF

Transformed Moment Goodness-of-fit

The Euclidean distance between the empirical standardized transformed moments and the expected center of the transformed moments (1/2,sqrt(1/12))

TQ-GOF

Transformed Quantile Goodness-of-fit

Sum the absolute differences, over 25 quantile points, between the empirical cumulative distribution function and the expected cumulative beta distribution function (uniform distribution function)

BR

Best rank of 8 single filter methods

The best rank value of 8 single filter methods (the highest rank value)

AR

Average of the top 2 ranks

The average of the best two rank values of 8 single filter methods

WAR

Weighted average of the top 4 ranks

The weighted average of the best four rank values of 8 single filter methods (weight = 4:3:2:1)