Struct ndarray_histogram::histogram::strategies::FreedmanDiaconis

pub struct FreedmanDiaconis<T> { /* private fields */ }

Robust (resilient to outliers) strategy that takes into account data variability and data size.

Let n be the number of observations and at = 1 / 4.

bin_width = (1 - 2 × at) × IQR × n^−1/3

The bin width is proportional to the interquartile range (IQR) from at to 1 - at and inversely proportional to cube root of n. It can be too conservative for small datasets, but it is quite good for large datasets. In case the IQR is close to zero, at is halved and an improper IQR is computed. This is repeated as long as at >= 1 / 512. If no IQR is found by then, Scott’s rule is used as asymptotic resort which is based on the standard deviation (SD). If the SD is close to zero as well, this strategy fails with BinsBuildError::Strategy. As there is no one-fit-all epsilon, whether the IQR or standard deviation is close to zero is indirectly tested by requiring the computed number of bins to not exceed max_n_bins with a default of u16::MAX.

The IQR is very robust to outliers.

Notes

This strategy requires the data

• not being empty
• not being constant
• having positive IQR

Implementations

impl<T> FreedmanDiaconis<T>

where T: Ord + Send + Clone + FromPrimitive + ToPrimitive + NumOps + Zero,

pub fn bin_width(&self) -> T

The bin width (or bin length) according to the fitted strategy.

Trait Implementations

impl<T> BinsBuildingStrategy for FreedmanDiaconis<T>

where T: Ord + Send + Clone + FromPrimitive + ToPrimitive + NumOps + Zero,

fn from_array<S>(a: &ArrayBase<S, Ix1>) -> Result<Self, BinsBuildError>

where S: Data<Elem = Self::Elem>,

Returns Err(BinsBuildError::Strategy) if improper IQR and SD are close to zero. Returns Err(BinsBuildError::EmptyInput) if a.len()==0. Returns Ok(Self) otherwise.

fn from_array_with_max<S>( a: &ArrayBase<S, Ix1>, max_n_bins: usize ) -> Result<Self, BinsBuildError>

where S: Data<Elem = Self::Elem>,

Returns Err(BinsBuildError::Strategy) if improper IQR and SD are close to zero. Returns Err(BinsBuildError::EmptyInput) if a.len()==0. Returns Ok(Self) otherwise.

type Elem = T

fn build(&self) -> Bins<T>

Returns a Bins instance, according to parameters inferred from observations.

fn n_bins(&self) -> usize

Returns the optimal number of bins, according to parameters inferred from observations.

impl<T: Debug> Debug for FreedmanDiaconis<T>

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.