T is for Table


tookie


T extends Pandas Dataframes with a collection of table manipulation methods as well as statistical, machine learning, financial and EDA methods.

For example it integrates Bootstrap ReSampling statistical methods (a.k.a Hacker Statistics)

Basic Usage

Create a plain data frame

>>> df = pd.DataFrame( {
    'user':['k','j','k','t','k','j']
    ,'period':['pre', 'pre', 'pre', 'pre', 'post','post'] 
    , 'kpi':[13,12,2,12,43,34]
    })
user period kpi
0 k pre 13
1 j pre 12
2 k pre 2
3 t pre 12
4 k post 43
5 j post 34


Filter the rows that have the value “post” and from that select the columns “user” and “kpi”

>>> t.select( t.where(df, "period", "post"), "user", "kpi")

Note that alternativelly we can also use the Pandas pipe operator to chain functions, with the same result I’ve renamed “pipe” to “p” when the T library is included:

(df
  .p(t.where, "period", "post")
  .p(t.select, "user", "kpi"))
. user kpi
0 k 43
1 j 34


Calculate the mean’s confidence interval. It includes plotting it.

>>> t.ci_mean(pd.DataFrame (np.random.normal(size=(37,2)), columns=['A', 'B']), 'A')
{'mean': -0.33, '95% conf int of mean': array([-0.64, -0.03])}

ci_mean

More Examples

https://github.com/al3xandr3/Data-Science-ipynb/blob/master/t%20is%20for%20table%20version2.ipynb

Run Tests

> cd "C:\path\my\projects\t"
> pytest



Project updates hosted in github: https://github.com/al3xandr3/T


Updated: