The bucket_comparator app creates an interactive visualization for a set of section data. It does statistical tests, decides whether two samples are from the same distribution and shows percentile breakdown.

Running

Installing Required Libraries

All required python packages are listed in bucket_comparator/requirements.txt. Go to bucket_comparator folder and install them using following command.

pip install -r requirements.txt

Bokeh Server

To view the app from a bokeh server, you need to run a bokeh server and point it at the bucket_comparator example directory. For example, from this directory, you can run:

bokeh serve .

Or from one directory level up, you can run:

bokeh serve bucket_comparator

Now navigate to the following URL in a browser:

http://localhost:5006/bucket_comparator

Data and Functionality

The data (bucket_comparator/data/data.csv) has the following colums: section, bucket, group_A, group_B.

There are 10 sections containing 201 buckets. Each bucket has count under two groups (group_A, group_B). When you run the app all sections will be available in a dropdown. You can browse the sections. For a selected section the app will do the following

• Show a count plot for all the buckets in two groups
• Do Kruskal–Wallis non-parametric test (https://en.wikipedia.org/wiki/Kruskal%E2%80%93Wallis_one-way_analysis_of_variance) and show whether group_A, group_B count values of the buckets are from the same distributon or not.
• Show test statistic and p-value of the Kruskal–Wallis non-parametric test
• Show percentiles (25, 50 and 75) of group_A, group_B counts for the buckets in a Table

You can also download the data of the selected section using "Download data" button.

Cheers!!!