CAUSALITY DISCOVERING

HOW STATISTICS CAN HELP

There is reason to believe that most Artificial Intelligence developers are confident that statistical data analysis (more specifically, analysis of the correlation between observed values) is an adequate tool for discovering causal relationships.

In this regard, the idea arose to invite readers to find causal relationships in the two cases represented by the respective datasets and then discuss the results.

The datasets are small enough to be imported into Excel or its equivalent: a thousand samples of four observed factors. The goal of the analysis is to discover causal relationships between these factors.

Correlation coefficients between the factors of the first data set:

1 2 3 4

1 1.0000 0.5889 -0.0386 0.0396

2 0.5889 1.0000 0.3504 0.4965

3 -0.0386 0.3504 1.0000 -0.0021

4 0.0396 0.4965 -0.0021 1.0000

Correlation coefficients between the factors of the second data set:

1 2 3 4

1 1.0000 0.9449 0.0299 0.0226

2 0.9449 1.0000 0.0480 0.1369

3 0.0299 0.0480 1.0000 -0.0523

4 0.0226 0.1369 -0.0523 1.0000

Original datasets (comma-separated format): data1.csv and data2.csv at https://github.com/mrabchevskiy/dataset

After two or three weeks, the results will be the subject of discussion.

We are waiting for your thoughts, decisions, and comments!