Chi Squared Test

A chi-squared test is a statistical test used to determine if there is a significant association between two categorical variables or if a sample data matches an expected distribution.

Considerations:

The script pulls the data from a .csv saved in the root of this repo. I've added an example input.csv, but do replace this with the data you want to perform the Chi Squared Test on. The significance level is set in run(), but this can be changed before running.

Here's how this data looks in a .csv format:

,responded,not_responded,total
treated,35,15,50
not_treated,26,29,55
total,61,44,105

The script returns a .json in the following format (with examples):

{'chi_square_statistic': 5.56, 'df': 4, 'critical_value': 9.488, 'p-value': 0.23, 'hypotheses': ['null: there is no relationship between x and y', 'alt: there is a relationship between x and y'], 'cell_data': [{'row_name': 'treated', 'row_total': 50, 'column_name': 'no_responded_to_treatment', 'column_total': 61, 'expected_cell_value': 29.05, 'observed_cell_value': 35, 'chi_squared_value': 1.22}, {'row_name': 'treated', 'row_total': 50, 'column_name': 'no_did_not_respond_to_treatment', 'column_total': 44, 'expected_cell_value': 20.95, 'observed_cell_value': 15, 'chi_squared_value': 1.69}, {'row_name': 'not_treated', 'row_total': 55, 'column_name': 'no_responded_to_treatment', 'column_total': 61, 'expected_cell_value': 31.95, 'observed_cell_value': 26, 'chi_squared_value': 1.11}, {'row_name': 'not_treated', 'row_total': 55, 'column_name': 'no_did_not_respond_to_treatment', 'column_total': 44, 'expected_cell_value': 23.05, 'observed_cell_value': 29, 'chi_squared_value': 1.54}]}

The data from the input.csv is cleaned and compiled new before being analysed. It relies on the data being in the format of the spreadsheet screenshot above. More rows can be added and more columns can be added and the headings can change. But total needs to remain as total and as the last row and last column. Cell A1 needs to be empty too.

I wrote this script to help a friend do this kind of work more quickly - so it's designed to suit her. But feel free to adapt this to your own data input when you clone the repo!

How to run:

1. Clone this repo [email protected]:sophiefitzpatrick/chi-squared-test.git
2. Run pip install -r requirements.txt
3. python run.py