I am doing an analytic exploratory analysis.

If the target is a continuous variable and the attributes are all categorical (discrete values), in order to know if exist any influence on the target from the each attribute I am doing the ANOVA-test like this:

fvalue, pvalue = stats.f_oneway(df[y], df[x])
pvalue < 0.5

If that condition is true, there is a dependency between variables.

For all variables I get true dependency with ANOVA, but the values of the correlation are between -0.1 and 0.2.

What kind of conclusion can I do with this?

Can I say for example that if in the variables with negative correlation, I get new values in order to make that correlation a high positive value, then the target will get a higher value?

Any ideas?


First of all as your target value is continuous so it's a regression problem.

If the correlation doesn't give desired value it means there exist no linear relationship between variable and target variable you should look into mutual information.

Also, while interpreting a relationship, one should be careful to not confound correlation and causality, because although a correlation demonstrates that a relationship exists between two variables, it does not automatically imply that one causes the other.

So if you increase the values to get positive correlation of that variable with target it might not necessarily increase the same as there might exist a confounder.

Mutual information versus correlation

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.