Sampling distribution of the sample mean. Sampling distribution: sampled from a probability distribution (that we typically don’t know) Sample mean: each sample we take consists of n data points and we take the mean of them. We plot all the means of samples we took, each sample with n data points Take sample with n data points, find mean of sample, plot it. The plot of sample means will be more normal as long as n the number of data points in each sample gets larger, having less skew and kurtosis (measure how normal our distribution is, skew is being shifted left or right and kurtosis is how pointy it is)
At 7:25, positive kurtosis should not look like that. it's tail should lower than normal on both side. The peak of negative kurtosis should not taller than normal.
@Khan Academy can you please clarify for me th following doubt. I got confused at 4:11 when you say ."And so when I click animated, what it's going to do is it's going to take FIVE SAMPLES FROM from this probability distribution function..." And you keep repeating in the video that it taes 5 samples.... shouldn't it be." And so when I click animated, what it's going to do is it's going to take ONE SAMPLE OF SIZE FIVE from this probability distribution function..." I'd appreciate some help on this. Just thought I understood the topic and then I hear the video's author talking about "taking five samples".
Jonathan Gordon literally search it up and judge it for yourself brainlet. with you asking that question, probably more knowledge than you will have within the next 5 years, and our district learnt it in the last year of hs
For the negative kurtosis example are the tails supposed to reach zero? Shouldnt they go on to infinity like the normal distribution and positive kurtosis?
Kurtosis is basically concerned about the shape of the peak of the frequency distribution, but ideally a thin tail can lead to a less pointy peak , that tail starts from infinity to infinity
But we used here only one random variable. I kinda find it weird. So I guess if it works for one with the given exp.val. and variance then we just add the outputs and still get the normal distribution?
The same size is 5 , he just took 5 sampling to illustrate he latter did 1000 sampling of the sample size of 5 and ploted the frequency distribution of mean of each of the 1000 sampling
It seems to me one needs to be sampling from a non pre-existing, infinite population in order for the CLT to work. For example, the generation of random variables from a population would need to come from something like a rolling of a die where there is no population "size" existing in advance. If the population had a size, like a population of people, then as n approached infinity, it would reach N and the distribution of the means would be a single value. Yes?
seems to me that a large enough sample size [or the cumulative results of many samples] if plotted would mirror the shape of the original data set ...the values would be different, but the relative frequencies at each position would be the same ...that's right isn't it?
Not exactly bro , the shape of the sample size will differ from the frequency of mean of samplings taken , we are more concerned that the distribution will look like a normal distribution as the size increase , but the mean of the frequency distribution will be close to the mean of the sample size
watching these at 5AM the morning of your statistics midterm at university is the best
thats about where i am rn
same
omg same bhahahaha
5:24am
Exactly me at %:07
You have been saving lives for the last 8 years, Keep doing like this!
Love this Khan guy he breaks it down so well.
thank you :) a slow student like me needs a patient tutor like you. thanks for helping me to understand
finally, a video about this that makes sense!
Apple Picker I know, right!
Sampling distribution of the sample mean.
Sampling distribution: sampled from a probability distribution (that we typically don’t know)
Sample mean: each sample we take consists of n data points and we take the mean of them.
We plot all the means of samples we took, each sample with n data points
Take sample with n data points, find mean of sample, plot it.
The plot of sample means will be more normal as long as n the number of data points in each sample gets larger, having less skew and kurtosis (measure how normal our distribution is, skew is being shifted left or right and kurtosis is how pointy it is)
Thank you
After 20 years of absence of study, just learning data science asynchronously, your videos are very helpful. thanks
10 years later and it's still relevant and helpful
"Skew is where there's few"
wow tbh thank you
wow now I understand it...
Thanks, it was nice to have a break from reading, yet not feel guilty about procrastinating.
Love how enthusiastic you are when teaching us
At 7:25, positive kurtosis should not look like that. it's tail should lower than normal on both side.
The peak of negative kurtosis should not taller than normal.
drinking game...drink when he says sample
Guess what! You're the best for me so far! Thank you so much
THANK YOU SO MUCH!
2021er here. Thank u
you are a blessing Sal Khan!
10 years old yet good video.
Thank you for all of your videos! They are so helpful!!!!
Thank you so much Sal
Thank you ALOT😭😭..I was completely lost before this video
so very much informative and intuitive. Thank You
this guy saves lives
Thanks!I FINALLY understood!!
So larger sample sizes would provide more accurate distributions
awesome video! THank you!
So well explained.
Thanks for the video! Might be stupid question but shouldn’t the sample size be noted as ”Sample n” instead of ”Sample N” on the CLT test tool?
Thank you 🙏
You are amazing!
I don't even understand what's going on but nice vid!
ty💙
If only my professors were half as good as you
Thanks so muchthese are great videos
Thank you so much
@Khan Academy can you please clarify for me th following doubt.
I got confused at 4:11 when you say ."And so when I click animated, what it's going to do is it's going to take FIVE SAMPLES FROM from this probability distribution function..."
And you keep repeating in the video that it taes 5 samples....
shouldn't it be." And so when I click animated, what it's going to do is it's going to take ONE SAMPLE OF SIZE FIVE from this probability distribution function..."
I'd appreciate some help on this. Just thought I understood the topic and then I hear the video's author talking about "taking five samples".
I am here in 2024 to study the sampling distribution of the sample mean. I came because I thought I was the only one who found it challenging.
I love khan academy
cheers mr khan
yall gotta get bro a new mic
Bravo!
You explain so well😍
outstanding!
Thanks
How difficult is the proof for the CLT? What level of math is needed to follow it?
Jonathan Gordon literally search it up and judge it for yourself brainlet. with you asking that question, probably more knowledge than you will have within the next 5 years, and our district learnt it in the last year of hs
does it pick certain number more/less often from the first graph, depending on the probability of the distribution?
MIDTERM TOMORROW!
i think he know everything about everything. he's basically god
Wow
I literally calculated the area of universe at a particular time by this methodology
how can you do that? the universe is 3d for an average guy who doesn't know relativity
you lying poo
That’s calculating the area of a sphere , you should get a Nobel price for lies
please, can any one tell me how like these videos manufactured ??
when comparing N=5 and N=25 I think you should look at the ratio kurtosis/sd (or sth like that) to infer about how close to the normal we are (?)
Perhaps the skew and kurtosis are better indicators for inference , it’s relation to standard deviation might be just for logical purpose I’m guessing
For the negative kurtosis example are the tails supposed to reach zero? Shouldnt they go on to infinity like the normal distribution and positive kurtosis?
Kurtosis is basically concerned about the shape of the peak of the frequency distribution, but ideally a thin tail can lead to a less pointy peak , that tail starts from infinity to infinity
@@victor_peral Thanks for bringing back my nightmares of university exams :D
@@iamb2348 can’t believe I replied from 6 years back 🤣😂
sampleception
nice
nice drawings 7.29
Whats the difference between median and mean?!
Median is when you got someone called Ian who is medical. Mean is when you are nasty to someone.
very bice thank you
😮😮you're the first onee
can anyone suggest some books
But we used here only one random variable. I kinda find it weird. So I guess if it works for one with the given exp.val. and variance then we just add the outputs and still get the normal distribution?
How to make samples in case of without replacement with size 4 can you know any trick please tell me
she's so great!
KHAAAAAAAAAAAAAAAAAAAAAAAAN
👍👍👍
He gets 5 samples and then averages them. What is the size of the original samples? 5 as well?
The same size is 5 , he just took 5 sampling to illustrate he latter did 1000 sampling of the sample size of 5 and ploted the frequency distribution of mean of each of the 1000 sampling
how can God not exist when everything works so beautifully in math and statistics
It seems to me one needs to be sampling from a non pre-existing, infinite population in order for the CLT to work. For example, the generation of random variables from a population would need to come from something like a rolling of a die where there is no population "size" existing in advance. If the population had a size, like a population of people, then as n approached infinity, it would reach N and the distribution of the means would be a single value. Yes?
skews are poopy like tails
seems to me that a large enough sample size [or the cumulative results of many samples] if plotted would mirror the shape of the original data set ...the values would be different, but the relative frequencies at each position would be the same ...that's right isn't it?
no
Not exactly bro , the shape of the sample size will differ from the frequency of mean of samplings taken , we are more concerned that the distribution will look like a normal distribution as the size increase , but the mean of the frequency distribution will be close to the mean of the sample size
This frickin class is skewed down👎. I HATE STATISTICS!!!!