Causality, Correlation and Regression

This video will explain you the commonalities and differences between the correlation, regression and the causality.
Causality means that there is a clear cause-effect relationship between two variables.
A common mistake in the interpretation of statistics is that when a correlation exists a causality is inferred.
There are two prerequisites for causality:
First, there is a significant relationship, that is, a Significant Correlation.
The second condition can be satisfied in two ways.
First, it is satisfied if there is a temporal ordering of the variables. So variable A was collected temporally before variable B.
Furthermore, the second condition can be fulfilled, if there is a theoretically founded and plausible theory in which direction the causal relationship goes.
If neither of the two is true, i.e. there is neither a temporal order nor can the causality be justified by a well-founded theory, then we can only speak of a relationship, but never of causality, i.e. it cannot be said that variable A influences variable B or vice versa.
More Information about Causality:
datatab.net/tu...
Regression Analysis: An introduction to Linear and Logistic Regression
• Regression Analysis: A...
Simple and Multiple Linear Regression
• Simple and Multiple Li...
Assumptions of Linear Regression
• Assumptions of Linear ...
Logistic Regression: An Introduction
• Logistic Regression: A...
Dummy Variables in Multiple Regression
• Dummy Variables in Mul...
Regression with categorical independent variables
• Regression with catego...
Multicollinearity
• Multicollinearity (in ...
Causality, Correlation and Regression
• Causality, Correlation...

[@dreamchaser4822]

You explain an often confusing concept very clearly in simple language which means you know exactly what you are talking about. And this is a common theme to all of your lectures. Thank you and with high respect.

[@datatab]

Thanks for the nice feedback!!!

[@errantvice7335]

This makes Elden Ring make soooo much more sense, thanks

[@andreashofmann6669]

A video about autocorrelation would be very cool. Great video and explanation!

[@datatab]

Many thanks, we will do it on our to do list!

[@hhy6620]

I am from Vietnam and in my country, The word causality is the same as regression and I have a lot of misunderstands but with this video with English, i have no misunderstands - sorry my English has no practicing

[@datatab]

Greate! Thanks for your feedback! Regards, Mathias

[@GemstoneActual]

Your English is good enough. Glad you beat the confusion. ;)

[@federicogarland272]

thank you, your videos are very helpful!

[@datatab]

Glad you like them!

[@mehmetb5132]

I am not sure just base on time difference we can conclude a causation! What if there is a third factor causing the initial two factors. Hypostatically, maybe child's mother's education is causing both child's age when she/he constructed her/his first sentence and her/his later school success instead? (en.wikipedia.org/wiki/Confounding)

[@datatab]

hanks for your comment! Yes, of course you have to be careful that the correlation does not actually come from a third variable! Regards, Hannah

[@muztosh3359]

thanks, i love you!!!

[@datatab]

: )

[@mazenhany2437]

good video as usual . I want to know more about causality and confounding

[@larissacury7714]

Thank youuuu!!!

[@datatab]

And again, thank you for watching : )

[@amanuelbeyene1460]

Good

[@datatab]

Thanks

[@mcchlizz]

Is instrumental variable as one of the method to deal with confounding variables to see the causality effect?

[@xco4555]

Base on your requirement, if a weather forecast for tomorrow is highly correlated with the result of tomorrow and take before the result that means the forecast is the cause of true weather ?

[@datatab]

That's a good point! But look at it this way, the forecast is not just given! The forecast is based on the weather today, with all the conditions that are known today, wind direction, temperature,.... and based on the weather information of today we make a prediction for the weather of tomorrow! So the weather of today is the reason for how the weather will be tomorrow (At least in part, I am not a metrologist) and since there is a correlation, you can predict the weather tomorrow with the weather today, this called weather forecast. If the weather was completely random and did not depend on the previous day, i.e. white noise, then of course no prediction could be made and the correlation would be zero. But if there is a correlation between today and tomorrow, it is clear in which direction the relationship goes, and you can make a prediction using a regression (or whatever). Therefore, whenever there is a time sequence and there is a correlation between the variables one can make a prediction. In the case of weather it is called a weather forecast, in the case of regression it is usually called a prediction. Just like in the stock market, if the price today or in the past is correlated with the price tomorrow, it is possible to make a prediction. Now, of course, it's a little twisted to say this prediction (or weather forecast) is the reason for the pries or weather tomorrow. The prediction was created exactly with the model that uses the correlation of „today“ and „tomorrow“. And the short answer: Since there is a correlation between the weather today and tomorrow, it is possible to predict the weather tomorrow with the weather today, this prediction is called weather forecast. Does that fit as an answer? Regards, Mathias

[@xco4555]

@@datatab So that prediction is Correlate or Causality ? I'm kinda new on this topic

[@datatab]

@@xco4555 If there is a correlation and the variables are separated in time, i.e. one event happens before the other, then we can speak of causality. If you want, you can then predict one variable with the other. If there is only a correlation, no temporal relationship and you don't know the direction of causality, you can't make a prediction. Simplified: correlation + causality = a prediction is possible. In the case of the weather: There is a relationship between the weather today and tomorrow, the causal direction is clear, therefore one can make a prediction. Cheers, Mathias

[@xco4555]

@@datatab I ask this question because I watch this clip kzbin.info/www/bejne/iaLckJVrhLicjtk which he said this relationship is correlation and your requirement said it causaulity

[@datatab]

@@xco4555 For what the requirement? One can have a correlation or not have, if there is a correlation one speaks of a relationship between the variables. But whether there is a causal relationship is not given purely by the correlation. This has to be investigated in more detail. When he talks in his video about education having an influence on salary and not the other way around, he probably does that under the assumption that education is before salary, so first you go to school and then you earn salary.

[@mpro9446]

this is the best explanation differentiating btwn correlation and regression I have yet found on the tube. thank you immensely.

[@putimaulida4880]

I'm still confused by the conclusions from some of the examples you made. Please enlighten me. So, 1. is the first example (age at which a child speaks her/his first sentences and later school success) IS a correlation? but why at the end of the video did you checklist YES on the causality requirement? 2. is the second example (intelligence and high school grade) IS a correlation?, and 3. is the last or third example (flies and body temperature) IS a causality?

[@captainseries913]

Remaining lecture of correction required in english language if possible thanks

[@datatab]

It will follow soon!

[@gezahegngashu3620]

The presentation is nice but I need the ppt

[@datatab]

Glad it was helpful! Thanks for your nice feedback! Regards Hannah