damn this is one of the best explanations I've seen for skewness and kurtosis this is kinda like activating cheatcodes to learning.. thank you so much for this explanation... much appreciated
Just because data has 0 skew and low excess kurtosis does not mean it has a normal distribution. There are an uncountable infinity of distributions and mixtures of them.
Such an amazing way of explanation! I learnt so much from this video :) Thanks so much !! Looking forward to more content from you, sir. Respect to you from India !! Liked and Subscribed !!
What does "has to" mean? You probably mean "to have similar kurtosis to a univariate Normal distribution? The "basic" formula for kurtosis will give a kurtosis of 3 for a normal, but MOST of the time formulas go ahead and subtract 3 to give the "excess kurtosis" compared to a normal distribution. Since that is the way 99% of stats formulas in programs calculate it, I am referring to "excess kurtosis" here.
Thanks for the great information sir... It clears all my doubt🤘... Just one thing I want to ask you, you have said that if the absolute value of skewness is less than 0.5 then the distribution is quite symmetric... But what about kurtosis??? You have told that if kurtosis is closer to zero then it's awesome but can you tell a significant range, by which I can easily decide whether I have to work on outliers or not... Again thanks for the information 🤘
Excellent!! Impressive!! I just wonder why do we need shapiro wilk test when we easily can determine normal distribution looking at the skewness and Kurtosis?
Thank you for nice presentation... I just want to ask about how to calculate standard error for skewness and kurtosis ether manually or by excel functions?
It is easy to Google this, but here is a Google result that explains it: estatistics.eu/what-is-statistics-standard-error-of-skewness-standard-error-of-kurtosis/
I'm a first time statistic student, I hope your videos will help me a lot. but if a Mean=1.373 833, Median=1452141, Skewness=0.0544 and Kuitosis=1.434581, what do they mean in the graph
Question! I have an exam tomorrow and there is one question in the mock exam that is driving me nuts! A researcher created a regression model and decided to use a logarithm on the "y-variable". What would be the reason to do so? A. The variable Y must have been positively skewed. B. The variable Y must have been negatively skewed. C. The variable Y must have been platykurtic. D. The variable Y must have been leptokurtic. Can anyone help me?
Taking the log can take some kinds of data that are positively skewed and make them more symmetric- e.g., there is a distribution called lognormal, that is skewed, but if you take the log it becomes a normal distribution. en.wikipedia.org/wiki/Log-normal_distribution
It does, but what every computer program actualy calculates if "Standardized" or "Excess" kurtosis compared to a normal distribution. Since it is extremely rare for people to actually calculate kurtosis manually, I am explaining what the numbers that Excel, R, SAS, Stata, SPSS, etc. will tell you. So, if you fed normally distributed data into one of these programs, you would get a value of zero.
You are correct- in the video I probably did emphasize the "peakness" too much because I was emphasizing what the words "platy" and "lepto" mean. In the video description I tried to add more on the "tailedness" idea to balance it out, and added (Read info below...) to the title to try to point people to that additional information. Maybe I should re-record this one so as to make the video better/more complete. I apologize if I added to your confusion.
BurkeyAcademy Thank you so much for clearing my doubt😊....no need to apologize, it's a big help in itself that you take time to share your knowledge with us😊❤
Of course I agree, which is why I say "Kurtosis measures whether the data has heavy tails (higher probability of outliers), or whether data is more concentrated in the center." This video is a brief, practical look, and I would be happy to make a more theoretical video if you like. However in practice, most of the time (in my experience) data with heavy tails also become more pointy looking, which is why the common terms "platykurtosis" (like a block) and "leptokurtosis" (narrow) came into being. For example, visually compare the shapes of the normal, t, and cauchy with excess kurtosis of 0, 6/(df-4) for df>4, and undefined (though thinking of it as infinite is justified). As you "squeeze" data from the middle out into the tails, it gets pointier. OF COURSE, the calculation of kurtosis does not directly measure pointyness, but it is a very common side effect. If you know of some common empirical cases where high kurtosis is coupled with LESS pointyness, I would love to learn about these cases! Thanks for the comment!
I was of the belief that kurtosis tells you nothing about the peak and only the tails? The tails in your explanation all seems quite similar. ”Kurtosis tells you virtually nothing about the shape of the peak- its only unambiguous interpretation is in terms of tail extremity, that is,either existing outliers (for the sample kurtosis) or propensity to produce outliers (for the kurtosis of a probability distribution)” (Westfall 2014)
If you read the extra information in the video description I clarify this a bit- "extremity" is the key word in your quote. I apologize for being a bit misleading in the video.
Did I just learn a concept in 11 mins which took my class teacher 4 hours to teach? Pinch me.
Thanks a ton.
Cool man! Thanks for letting me know!
Thanks for explaining so clearly. I think you are the best Stat tutor on youtube.
damn this is one of the best explanations I've seen for skewness and kurtosis
this is kinda like activating cheatcodes to learning.. thank you so much for this explanation... much appreciated
love this basic, down to earth explanation … now I get it … 100 thumbs UP for you sir.
now I can finally go back to completing my data science assignment, thankyou so much!
My god, I perfectly understood your explanation, spoken like a human... unlike my statistics prof...
After 5 long years.... i finally understood. thanks SirJi ♥
I'm agree whole heartedly. This is the BEST teacher of math and statistics out there! Plus, the gal who posted this comment is good looking!!!
better than my uni days, a million times better
Great video. The explanations with examples are so clear and easy to understand! Thanks!
Tony Stark teaching me statistics haha!
Thanks for this! I couldn't find any other videos or much information on what the kurtosis values ACTUALLY MEANT!
Awesome video! Super helpful for a grad student trying to learn how to analyze data
Skewness :
In Between -0.5 to 0.5 : Symmetry
>0.5 to 1 : Positive skewed
>1 : too Much skewed to positive side
-1 to -0.5 : Negative skewed
Just because data has 0 skew and low excess kurtosis does not mean it has a normal distribution. There are an uncountable infinity of distributions and mixtures of them.
So what should be ideal number for skewness and kurtosis to follow normal distribution.
One of the best videos ever! Thanks a lot!
Thank you so much. This helped me to understand how to approach numerous variables in my data set.
great explanation. making sense for me even without any background knowledge.
Thank you! The examples were clear and easy to understand.
Amazing and very simple explanation 👍🧐 thank you!
so thankful sir... very easy to learn from your presentation
Thank you, thank you, thank you!!
Such an amazing way of explanation! I learnt so much from this video :) Thanks so much !! Looking forward to more content from you, sir. Respect to you from India !! Liked and Subscribed !!
Thanks niceguy! Glad you liked it.
Hello !! Thank you so much, sir! It is my privilege :)
Perfect Explanation. Thanks Burkey :)
Dear sir, from an another website I have read, kurtosis has to approach 3? It's different from what you have told in 5:48.
What does "has to" mean? You probably mean "to have similar kurtosis to a univariate Normal distribution? The "basic" formula for kurtosis will give a kurtosis of 3 for a normal, but MOST of the time formulas go ahead and subtract 3 to give the "excess kurtosis" compared to a normal distribution. Since that is the way 99% of stats formulas in programs calculate it, I am referring to "excess kurtosis" here.
way of explanation is super.give us more subject if possible.
Made it easy to understand. Thank you!
Thank you so much for this video. Love your teaching style!
Thanks for the great information sir... It clears all my doubt🤘...
Just one thing I want to ask you, you have said that if the absolute value of skewness is less than 0.5 then the distribution is quite symmetric... But what about kurtosis??? You have told that if kurtosis is closer to zero then it's awesome but can you tell a significant range, by which I can easily decide whether I have to work on outliers or not...
Again thanks for the information 🤘
Thanks a lot for the explanation.
Excellent!! Impressive!! I just wonder why do we need shapiro wilk test when we easily can determine normal distribution looking at the skewness and Kurtosis?
Because no observed sample from a normal distribution will have exactly the properties is was drawn from.
Now I can understand my data distribution. Thanks a lot
Thanks for this video. It was really helpful.
Is there any video for 'Bassel's correction'?
No, but I could do that...
very great explanation
Why do we calculate standard deviation by using mean? Namely, why dont we use mode instead of mean?
Thank you this has been very helpful
THanks for the video. I really like it.
Ohh so great this video. Thanks so much!
Which playlist does this video belong to?
It wasn't in a playlist, so I added it to this one on "Numerical Descriptive Statistics: ua-cam.com/play/PLlnEW8MeJ4z4YdizTw_wV4HThhHJ2zp0F.html
@@BurkeyAcademy thank you :)
@@BurkeyAcademy thank you :)
Amazing explainatin thanks !!!
Thank you very much ! This was very useful !
Fantastic video. Great communication.
very well explained...thank you :)
Kurtosis has to do with relative frequency of outliers not "pointiness."
This helped so much! Thank you!
Thank you for nice presentation... I just want to ask about how to calculate standard error for skewness and kurtosis ether manually or by excel functions?
It is easy to Google this, but here is a Google result that explains it: estatistics.eu/what-is-statistics-standard-error-of-skewness-standard-error-of-kurtosis/
Thank you for fast response.. Actually I tried that by google but for some reason, any link with that issue did not opened, also yous !! Thanks again.
Happy to help!
So, if a histogram has, say, 4 bars and all have same frequency, it will be symmetric graph, not left/right skewed? Also what will be it's kurtosis?
Thanks Buddy: Subscribed
Is it just me or he sounds like Mr Stark? (Have I watched too much Marvel Movies? :o )
Now this is new... I might like this suggestion! :) People usually say "Tom Hanks".
So when an age distribution has a positive kurtosis, then it means that the distribution is pointy, okay.
I might be pointy, or could have other shapes with more outliers than a normal distribution.
Nice and easy to understand.Thanks
Very well explained, thank you very much
I'm a first time statistic student, I hope your videos will help me a lot. but if a Mean=1.373 833, Median=1452141, Skewness=0.0544 and Kuitosis=1.434581, what do they mean in the graph
almost normally distributed data , i think median value is not right
Very cool video, thank you!
Great video
can curtosis be exactly 0?
Yes, it is possible (software usually reports excess kurtosis which is zero for a normal distribution)
Love it!
Question! I have an exam tomorrow and there is one question in the mock exam that is driving me nuts! A researcher created a regression model and decided to use a logarithm on the "y-variable". What would be the reason to do so?
A. The variable Y must have been positively skewed.
B. The variable Y must have been negatively skewed.
C. The variable Y must have been platykurtic.
D. The variable Y must have been leptokurtic.
Can anyone help me?
Taking the log can take some kinds of data that are positively skewed and make them more symmetric- e.g., there is a distribution called lognormal, that is skewed, but if you take the log it becomes a normal distribution. en.wikipedia.org/wiki/Log-normal_distribution
You are a hero among men, thank you so much!
I thought normal distribution had a kurtosis of 3?
It does, but what every computer program actualy calculates if "Standardized" or "Excess" kurtosis compared to a normal distribution. Since it is extremely rare for people to actually calculate kurtosis manually, I am explaining what the numbers that Excel, R, SAS, Stata, SPSS, etc. will tell you. So, if you fed normally distributed data into one of these programs, you would get a value of zero.
I have a doubt..... some people say kurtosis is measure of 'tailedness' , but you mentioned it measures the peak? So confused!😅
You are correct- in the video I probably did emphasize the "peakness" too much because I was emphasizing what the words "platy" and "lepto" mean. In the video description I tried to add more on the "tailedness" idea to balance it out, and added (Read info below...) to the title to try to point people to that additional information. Maybe I should re-record this one so as to make the video better/more complete. I apologize if I added to your confusion.
BurkeyAcademy
Thank you so much for clearing my doubt😊....no need to apologize, it's a big help in itself that you take time to share your knowledge with us😊❤
Really, It was awesome.
data set link is not working kindly upload new link
I just tested the link, and it is working fine.
how did you get the numbers in age and etc...? please reply ASAP
The link is in the description of the video.
Great vid. I've subscribed, will check out the rest of your content later.
its not pointy what kurtosis tells ,its the outlier in the datasets which makes high kurtosis
Of course I agree, which is why I say "Kurtosis measures whether the data has heavy tails (higher probability of outliers), or whether data is more concentrated in the center." This video is a brief, practical look, and I would be happy to make a more theoretical video if you like. However in practice, most of the time (in my experience) data with heavy tails also become more pointy looking, which is why the common terms "platykurtosis" (like a block) and "leptokurtosis" (narrow) came into being. For example, visually compare the shapes of the normal, t, and cauchy with excess kurtosis of 0, 6/(df-4) for df>4, and undefined (though thinking of it as infinite is justified). As you "squeeze" data from the middle out into the tails, it gets pointier. OF COURSE, the calculation of kurtosis does not directly measure pointyness, but it is a very common side effect. If you know of some common empirical cases where high kurtosis is coupled with LESS pointyness, I would love to learn about these cases! Thanks for the comment!
@@BurkeyAcademy i understand the now more about the pointy ness as we are trying to squeeze the shoulders and spread the data on tails.
Thank you so much. !
superb👌
Very informative. Thanks
4:00
I was of the belief that kurtosis tells you nothing about the peak and only the tails? The tails in your explanation all seems quite similar. ”Kurtosis tells you virtually nothing about the shape of the peak- its only unambiguous interpretation is in terms of tail extremity, that is,either existing outliers (for the sample kurtosis) or propensity to produce outliers (for the kurtosis of a probability distribution)” (Westfall 2014)
If you read the extra information in the video description I clarify this a bit- "extremity" is the key word in your quote. I apologize for being a bit misleading in the video.
Mr. Stark is that you
Much better
Excellent
sir cannot find the raw data file for child health and development study and no link in the description.
Thanks for letting me know, and sorry about that! I added the link now!
thank you... so much sir.
BurkeyAcademy
thx very much
You sound like Tony Stark... Holy
It's never peakedness but tailedness
I am Iron Man!
You sound like Iron Man :D
Just sayin'. Your voice sounds like Iron man's 7u7
Great explanation! Thank you