I am looking for a tool or Python library, to analyse data and (automatically) find correlations between several variables (more than 2, numerical or not).
These are the kind of things I would like to get from a data set having 10+ variables:
- The correlation between each of the variables (1vs1) (Excel does it)
- Details about this correlation (e.g. A and B have a positive correlation when A>70)
- The correlation between several variables (e.g. A and B have a positive correlation, only when C>70)
- Correlation with non numerical values (e.g. people's sport performance and weight have a negative correlation (the fatter, the less sportive) for weight>70, except in Japan where this correlation is positive because the sport is Sumo!
- The answer to the kind of question : how to maximise variable F ? And the answer would be : take the records having A=USA and having B,D positively correlated
I have created a small excel file with a data set with this example. And I have done a chart which intuitively helps to guess these rules. http://s000.tinyupload.com/?file_id=09217280893718830417 But What if I had a big data set? What tool could I use?