Python module to compute running statistics over data, like when measuring timings from a stream.

Properties:

• count
• min,max,mean
• std and variance
• kurtosis and skewness
• merge of two livestats preserving statistics
• arithmetic operation over stat: + - * /
• standardization of the stat: uses mean and std to compute offset and scaling and then applies to the data updating the other statistics
• minmax normalization to [0,1]: uses mean and std to compute offset and scaling and then applies to the data updating the other statistics

Normality tests:

• jarque_bera
• kurtosis and skewness

The main class is LiveStat to which data can be appended with append(x). For incremental values the DeltaLiveStat provides an easy to use helper.

Usage:

from livestat import LiveStat,DeltaLiveStat

x = LiveStat("optionalname")
x.append(10)
x.append(20)
print x # count is 2

x = DeltaLiveStat("dt")
x.append(10)
x.append(20)
print x # count is 1 containing the difference

#also from array
x.extend([10,20,30,40,50])

Extra Features:

# the LiveStat objects can be combined for example when performing over different data Windows or in a multiprocessing environment
x.merge(y) # now x contains the merge of the statistics

# the LiveStat object can be multipled by scalar or translated, for the objective of performing some unit transformation. All the measures are transformed appropriately
x + 5
x * 5

• numpy support for fast forwarding append, and for vectorial statistics

This project is maintained here: https://github.com/eruffaldi/pylivestat

The faststat package is similar:

https://pypi.python.org/pypi/faststat/
https://github.com/doublereedkurt/faststat/

More sophisticated features can be found in scipy.stats https://docs.scipy.org/doc/scipy-0.14.0/reference/tutorial/stats.html