Kurtosis is critically important and it's not typically discussed in entry- or even intermediate-level statistics courses. So happy to see you covering it here, your content deserves an Olympic gold medal. Thank you so much
As someone who is currently transitioning into Data Science from an Econ background, I appreciate the intuitive examples you use in your videos! Thanks and keep up the great work!
Just started learning about kurtosis for a finance course and am very happy to have found your video! Very helpful. I love how you break things down so that they're easy and intuitive to understand.
I first learned about kurtosis in my high school research class - it was a stat we looked at for our project, but I really didn't know what it was aside from being a weird word... Thank you for the explanation. This is a great, well-needed video!
Very good. A slight clarification though: heavy tails does not necessarily mean "a lot of outliers." A single observation that is 7 standard deviations from the mean is enough to indicate a heavy tail. After all, such an observation would not happen, for all intents and purposes, under normality. The tails of heavy tailed distributions, while much higher than normal tails, are still very close to zero.
Kurtosis does not measure peakedness or flatness. Perfectly flat-topped distributions can have very high kurtosis, and infinitely peaked distributions can have very low kurtosis. Examples of such distributions are given on the current Wikipedia page.
One thing that kills me is when, like in the video example, the blue distribution is skewed to the left, my brain wants to understand as the other way around, e.g, 'to the right', because that is what the figure looks like, as if it was 'pending to the right'. D:
9:16 But you said that the standard deviations of both distributions is the same, correct? How is that possible that the number of outliers in both distribution differs, YET they still both have the same standard deviation? Thanks!
Problem with these higher moments is with their sensitivities to outliers. How much more data we need to "reliably" estimate these (ever-increasing) higher moments?
Thank you for this viseo. For some reason never thought about including kurtosis in my descriptive statistics analysis. Do you think you can do (or maybe you already did) vidoe on advanced statistics analysis: cdf, etc. More from the perspective what else can be added to it. Thank you!
I think the full distribution in visual form is the best for non-technical audiences. Then you don't have to dumb down and you don't have to explain it much either. If graphs are forbidden for some odd reason, maybe "buckets" can be a good way. Like 20% of the students had less than 30p, 30% in range 30-50p etc. Honestly even standard deviation is tricky to understand even for technical people. Ask someone who "should know" to explain - on the spot - the difference between standard deviation and the mean deviation.
Thank you for the explanation! Can you further clarify the difference between standard deviation and kurtosis? They feel similar as they relate to outliers to some degree, at least when you explained it in your example. Thanks!
![USEFUL OR TRASH? Kurtosis and Skewness [WATCH THIS, BEFORE YOU USE]](/img/1.gif)
Kurtosis is critically important and it's not typically discussed in entry- or even intermediate-level statistics courses. So happy to see you covering it here, your content deserves an Olympic gold medal.
Thank you so much
thanks for the words! Kurtosis is one of my favorite topics haha
@ritvikmath Yes, I ❤️ that word! It sounds like a disease... I am going to stay home today because I have the kurtosis! 😉
As someone who is currently transitioning into Data Science from an Econ background, I appreciate the intuitive examples you use in your videos! Thanks and keep up the great work!
+1
Glad it was helpful!
Just started learning about kurtosis for a finance course and am very happy to have found your video! Very helpful. I love how you break things down so that they're easy and intuitive to understand.
Glad it was helpful!
which course
I first learned about kurtosis in my high school research class - it was a stat we looked at for our project, but I really didn't know what it was aside from being a weird word...
Thank you for the explanation. This is a great, well-needed video!
After 10 years of education, I realized you are a super statistician and teacher!!!! Thanks
Thank you for explaining the topic in such a simple manner, this topic initially was quite difficult. Thanks!
I have a bachelor's and Masters degree in statistics and I can say this is the best explanation so far on these concepts
9:52 correction: kurtosis of a standard normal distribution is 3. the excess kurtosis is then whatever the kurtosis is minus 3.
Lucky to have bumped on this video. Learnt a lot from it. Thanks for your approach. It's true most lectures avoid these.
I just started studying statistics and this video was really useful and accessible, thank you!
Hi, thank you sooo much for the explanation! You really brought out the most important characteristics and meaning of each metrics. :)
Glad it was helpful!
I learned more than any of my uni courses in this video
Thank you so so much
Very good. A slight clarification though: heavy tails does not necessarily mean "a lot of outliers." A single observation that is 7 standard deviations from the mean is enough to indicate a heavy tail. After all, such an observation would not happen, for all intents and purposes, under normality. The tails of heavy tailed distributions, while much higher than normal tails, are still very close to zero.
Very useful especially the example the application of kurtosis and skewness on growth and inequality
Thanks!
Your videos are really amazing. Thank you so much for the intuitive and motivating examples. They are true gold.
Thank you. Well done job of fleshing out the topics.
My man- u r my savior and a legend.❤
I really want to know more about Kurtosis! Please make a video about it!😊
Thanks for the suggestion!
I never understood moments of a distribution.
12:21 hit me like a brick when I realised.
It’s quite a strange concept unless properly motivated
same here - there’s a clarity to the way this is outlined here that is just crystalline. subbed
very well explained , clear and Crisp
Skewness measures the degree of asymmetry of the distribution, while Kurtosis measures the degree of peakedness and flatness of a distribution
Kurtosis does not measure peakedness or flatness. Perfectly flat-topped distributions can have very high kurtosis, and infinitely peaked distributions can have very low kurtosis. Examples of such distributions are given on the current Wikipedia page.
9:51
A Normal distribution has a kurtosis of 3 not zero.
Excess kurtosis for normal distribution is zero.
Pls correct me if I'm wrong.
yes you are right. I missed the word "excess" there.
This is really great!
I have a BA in Stats and can confirm the accuracy of the video's title.
Best explanation on this topic :)
12:54 Wow, I didn't know about moments - except for "moment-generating functions".
Awesome explanation. Bravo!
Many thanks!
11:29 What about the median?
Which is better - mean or median? Could you do a video on this?
Thanks!
Excellent summary, thank you
One thing that kills me is when, like in the video example, the blue distribution is skewed to the left, my brain wants to understand as the other way around, e.g, 'to the right', because that is what the figure looks like, as if it was 'pending to the right'. D:
9:16 But you said that the standard deviations of both distributions is the same, correct?
How is that possible that the number of outliers in both distribution differs, YET they still both have the same standard deviation?
Thanks!
Very much needed. Thank you
Great example sir , I Understood
"Look at full distr." + "plot that full distribution and think about..."
The best descriptor of a set is the set :-)
Pretty much!
Loved this video!
Can you do a similar video with examples on other moments and things like statistical entropy? 😁
Great suggestion!
Great video! Thank you.
Problem with these higher moments is with their sensitivities to outliers. How much more data we need to "reliably" estimate these (ever-increasing) higher moments?
That is an interesting angle to look at it!
Can you explain how to.calculate skewness and kurtosis?
Can you Please explain how to measure skewness and kurtosis numerically? 👏❤️
Thank you for this viseo. For some reason never thought about including kurtosis in my descriptive statistics analysis.
Do you think you can do (or maybe you already did) vidoe on advanced statistics analysis: cdf, etc. More from the perspective what else can be added to it. Thank you!
I think the full distribution in visual form is the best for non-technical audiences. Then you don't have to dumb down and you don't have to explain it much either. If graphs are forbidden for some odd reason, maybe "buckets" can be a good way. Like 20% of the students had less than 30p, 30% in range 30-50p etc. Honestly even standard deviation is tricky to understand even for technical people. Ask someone who "should know" to explain - on the spot - the difference between standard deviation and the mean deviation.
You are the best!
Thanks!
Thank you for you video.
You are welcome
Very well done
Cheers pal!
kurtosis of normal distribution is 3. Excess kurtosis is amount exceeding 3.
👏 👏 👏 Very well done!
Thank you! Cheers!
Thank you for the explanation! Can you further clarify the difference between standard deviation and kurtosis? They feel similar as they relate to outliers to some degree, at least when you explained it in your example. Thanks!
Kurtosis puts more weight to observations that are far away from the mean.
What are the meaning of those symbols in the formula
Thankss!
CFA level 1 brought me here
THANK YOU, in my case for describing Kurtosis.
Of course!
Good old pen and paper.
Wow!
Congratulations ♥, You got a new Subscriber!.
@ritvikmath, You explained it in very intuitive way.