Pairwise Compare

This tool allows the user to compare groups of data specified in a tidy pandas dataframe with ease. In this repo the capabilities of the PairwiseCompareManager are shown through examples using a dataset from the nf1_schwann_cell_painting_data. These examples can be found in the docs. Although, most of the development efforts can be found in the src folder. Users should almost exclusively interact with the PairwiseCompareManager, however, there may be rare exceptions. If you choose to interact with another component of the tool, then there will be less input validation safeguards available.

Installation

You can install this tool with either of the following commands:
pip install pairwise-compare
uv add pairwise-compare

Once installed you can use subtools such as the PairwiseCompareManager with:
from pairwise_compare.comparison_tools.PairwiseCompareManager import PairwiseCompareManager
and a comparator with, for example:
from pairwise_compare.comparators.PearsonsCorrelation import PearsonsCorrelation

Development

See CONTRIBUTING.md for how to run the tests and make a release.

Data Formats

When passing arguments to the PairwiseCompareManager you can specify the columns that remain the same in each group-to-group comparison, and the columns that will be different in these comparisons. These columns are parameterized by _same_columns and _different_columns, respectively. The column values in these columns uniquely define each group. During pairwise comparisons of groups, all of the column values of the columns specified in _same_columns will be the same between both groups compared for all paired combinations of groups. Likewise, all of the column values of the columns specified in _different_columns will be different between both groups compared for all paired combinations of groups.

One of the following column arguments conditions must be satisified when using the PairwiseCompareManager:

1. _same_columns must include at least one list element if _different_columns has less than two list elements.
2. _different_columns must contain one or more list elements.
3. _same_columns and _different_columns should not contain any of the same columns.

Limitations

1. Input validation is enforced in the PairwiseCompareManager and PairwiseCompare classes.
2. Additional column values are not tracked aside from columns used to compare groups (_same_columns, _different_columns).
3. Output and input python data structures are limited.
4. All of the data, in the supplied pandas dataframe, is used to compute comparisons.

Extensibility

This tool compares features between any two groups of a tidy pandas dataframe. To incorporate additional comparators for making comparisons, you must introduce the functionality as a class and inherit from the Comparator class (See Comparator.py for details).