Home

Star us on GitHub!

Star

Home

OnlineStats is a Julia package for statistical analysis with algorithms that run both online and in parallel. Online algorithms are well suited for streaming data or when data is too large to hold in memory. Observations are processed one at a time and all algorithms use O(1) memory.

Installation

import Pkg
Pkg.add("OnlineStats")

Basics

Every stat is <: OnlineStat{T}

(where T is the type of a single observation)

julia> using OnlineStats

julia> m = Mean()
Mean: n=0 | value=0.0

julia> supertype(typeof(m))
OnlineStat{Number}

Stats can be updated

fit!(stat::OnlineStat{T}, y::S) will iterate through y and fit! each element if T != S.

julia> y = randn(100);

julia> fit!(m, y)
Mean: n=100 | value=0.0662854

Stats can be merged

julia> y2 = randn(100);

julia> m2 = fit!(Mean(), y2)
Mean: n=100 | value=0.0572355

julia> merge!(m, m2)
Mean: n=200 | value=0.0617605

Stats have a value

julia> value(m)
0.06176045359627141

Collections of Stats

Series

A Series tracks stats that should be applied to the same data stream.

y = rand(1000)
s = Series(Mean(), Variance())
fit!(s, y)
Series
  ├── Mean: n=1000 | value=0.492071
  └── Variance: n=1000 | value=0.0793655

FTSeries

An FTSeries tracks stats that should be applied to the same data stream, but filters and transforms (hence FT) the input data before it is sent to its stats.

s = FTSeries(Mean(), Variance(); filter = x->true, transform = abs)
fit!(s, -y)
FTSeries
  ├── Mean: n=1000 | value=0.492071
  └── Variance: n=1000 | value=0.0793655

Group

A Group tracks stats that should be applied to different data streams.

g = Group(Mean(), CountMap(Bool))
itr = zip(randn(100), rand(Bool, 100))
fit!(g, itr)
Group
  ├── Mean: n=100 | value=0.124212
  └── CountMap: n=100 | value=OrderedCollections.OrderedDict(false=>50,true=>50)

Additional Resources