Data.TDigest.Postprocess

TDigest postprocessing functions.

These are re-exported from Data.TDigest module.

Histogram

histogram :: TDigest comp -> Maybe (NonEmpty HistBin)

Calculate histogram based on the TDigest.

data HistBin

Histogram bin

Constructors

HistBin

Fields

• hbMin :: !Double

  lower bound

• hbMax :: !Double

  upper bound

• hbValue :: !Double

  original value: (mi + ma) / 2

• hbWeight :: !Double

  weight ("area" of the bar)

• hbCumWeight :: !Double

  weight from the right

Instances

Show HistBin

Methods

showsPrec :: Int -> HistBin -> ShowS

show :: HistBin -> String

showList :: [HistBin] -> ShowS

Quantiles

median :: TDigest comp -> Maybe Double

Median, i.e. quantile 0.5.

quantile :: Double -> TDigest comp -> Maybe Double

Calculate quantile of a specific value.

Mean & variance

As we have "full" histogram, we can calculate other statistical variables.

mean :: TDigest comp -> Maybe Double

Mean.

>>> mean (tdigest [1..100] :: TDigest 10)
Just 50.5

Note: if you only need the mean, calculate it directly.

mean' :: NonEmpty HistBin -> Double

Mean from the histogram.

variance :: TDigest comp -> Maybe Double

Variance.

variance' :: NonEmpty HistBin -> Double

Variance from the histogram.

CDF

cdf :: Double -> TDigest comp -> Double

Cumulative distribution function.

Note: if this is the only thing you need, it's more efficient to count this directly.

icdf :: Double -> TDigest comp -> Maybe Double

Alias of quantile.

NonEmpty

histogram' :: NonEmpty (Mean, Weight) -> NonEmpty HistBin

Histogram from centroids

quantile' :: Double -> Weight -> NonEmpty HistBin -> Double

Quantile from the histogram.

Debug

validateHistogram :: Foldable f => f HistBin -> Either String (f HistBin)

Validate that list of HistBin is a valid "histogram".