NOTE: At 7:49 I meant to say Drug C instead of Drug A. Support StatQuest by buying my book The StatQuest Illustrated Guide to Machine Learning or a Study Guide or Merch!!! statquest.org/statquest-store/
Hey josh i suggest you to write a book on everything you explained in your ML and prob&stats playlist. It will surely be a triple baaammm in the community. And also going to be best seller.
If only there was some nice feature to add quick, like, _annotations_ to videos to point out small stuff like that. Jeez, my hidden genius is scary =_= /s @RUclips
You know what I love about your channel Josh? You've made a real effort to get to the bottom of these concepts. It's beautiful to see such attention to detail and love for a subject.
I literally watched your Statistics Fundamentals playlist in 2 days. It was too much of wisdom. I need to sit back, go through my notes again. Crazy crazy learning. Thank you so much for putting so many efforts that you made them look so effortless. Thank you, sir!
Josh's voice is so cathartic to listen to and there's a perfect balance of learning at a great pace with minor sarcasm/banter that makes these videos so easy to binge and learn from. Thank you sir.
Will take my comps exam for grad school this coming saturday. Been using your videos to brush up on quanti methods. As a visual person, is so helpful. Thank you so much! Wish me luck! :D
Josh... and all of you friendly folks of the genetics department. Thank you guys SO MUCH for your great work. Thanks to you I learned so much in an intuitive way.
This is so good! Thank you so much! I tried watching other videos about this but none could explain it well as you do nor keep my attention long enough.
Thanks Mr Josh for the wonder video.... You are taking lots of effects to make the video....... your clarity of explanation is really good ..... possible do it for anomaly detection in time series ........Thanks a lot for making more videos.
Of course like! Awesome lectures! Explained very clearly and easily - the best I found. Unfortunately, I did not find anything 'clearly explained' at the channel about Markov models.
C, Aminor, G, C. BAM! :) (NOTE: They might sound a little different played on a normal guitar. The notes will be the same, but the order they are played might be different.)
Thanks Josh for the intuition & crystal clear explanation for Hypothesis Testing - one of the toughest topics to explain to someone. TRIPLE BAM !!! @statquest Your explanation is better than any other resources FREE or PAID on the whole Internet. TRIPLE BAM !!! @statquest
Your stat fundamental series realy helped me understand the basics of hypothesis testing, huge thanks for that! I searched for a video explaining the mathematical logic behind z and t test but found no one that realy explained the Details in an understandable way. Would you mabey consider making a video based on all the stuff you allready explained but connecting the dots and other neccecary details for z and t tests? Would be great if you'd think about it. (Unless you allready made one and I just didnt found it^^)
You know, one day I would love to do a proper video on t-tests. For now, however, the only thing I have is a video that shows you that the logic behind t-tests is the same exact logic behind linear regression, which is also the same exact logic behind ANOVA and a bunch of other tests (some of which don't even have names). For details, first learn about linear regression... ruclips.net/video/nk2CQITm_eo/видео.html and then learn how it naturally extends to t-tests... ruclips.net/video/NF5_btOaCig/видео.html
Josh- what about scientific hypotheses regarding forensic cases- where the null necessarily must point to 'not guilty' and therefore a fingerprint match null would be "different" and the alt "not different"!
I'm not 100% sure I understand your question, but you could imagine using DNA and saying "the null is that these two samples are the same" and then showing that there are enough differences to reject that hypothesis.
I am trying to calculate the sample size to look up for five alleles frequencies in my population and after searching I found this equation: Sample Size = [z2 * p(1-p)] / e2 ,but someone just suggested that it should be calculated using the hardy equation!!! that's why I have asked if it can be used for sample size calculation!
Hello! Do you think that I could get a basic notion of inferential statistics after watching the whole "Statistics Fundamentals" videos? If not, which ones do you recommend me to watch? :) BTW this videos changed my belief that I could actually understand something with numbers involved
This is the first time I've heard of the null hypothesis defined as there being no difference between observations in test results. I always interpreted the null as the observations being due to randomness instead of being due to some causal relationship. This definition would be more in line with the p-value being the probability that the observations actually are random. Of course, I could just be super confused :P
Another way to think of the null hypothesis is that all of the observations were randomly selected the same distribution (instead of two, distinct distributions). If all the observations came from the same distribution, then, in theory, there should be no, or relatively little, difference bean their estimated means.
There are ways to determine the number of experiments needed and I describe them in these videos: ruclips.net/video/Rsc5znwR5FA/видео.html and ruclips.net/video/VX_M3tIyiYk/видео.html
Great video however, I have a question. Throughtout the video, you used the phrase "random differences" for factors like Better diet, more exercise, (or may be some genetic mutation) which affect recovery time but are these factors really random ? These factors may be unknown to the experimenter but is it a good idea to consider them random ? I mean the phrase "Drug A may perform better than Drug B due to chance" doesn't make sense considering there would be deterministic reasons with mechanical explanations which would explain the action of the drugs
I agree, statisticians use words like "random" and "error" in strange ways. However, in the language of statistics, "random" means "things we didn't factor into the model". Unfortunately, that's just the way they use the term and there's not much we can do about it.
Great video, thanks! I completely understand what you said about us only being able to reject or fail to reject the null hypothesis, but I have seen many published papers stating their main hypotheses to be "there is no difference between groups". They then state in the discussion "the results support our hypothesis" when the tests find no difference and claim that the intervention is equally helpful for different age groups, for example. Is this problematic? Should the peer review have said something? Should they have switched the null hypothesis and the alternative hypothesis somehow because the type I error becomes the type II error and vice versa? What would be the algorithm for doing such research when they want to prove that the intervention is as effective as the other one?
Without knowing all of the details, I would say it is very odd to say that the main hypothesis is the null and that a large p-value supports that. The reason it's odd is that there could be a million reasons why the p-value was large, some of which have nothing to do with the treatment (for example, it could be that someone just wrote the data down incorrectly).
This is awesome...can u please make video on Skewness and Kurtosis ... explaining it's concept and practical application.... Thanks in advance for reply
Hi Mr Starmer, the lecture overall is incredible. Im just stuck at one point where I think the explanation flow is not as smooth as others. I know, BY COMMON SENSE, that the recovery time difference when giving drug C and D the 2nd and 3rd time is different, but not enough to reject hypothesis 1. Can u give me perhaps a more detailed and scientific explanation about why we simply are not certain to claim hypothesis 1 is incorrect, even though it is different then the following 2nd and 3rd time? Like how do we know if it is very likely that the hypothesis can be rejected?
OK. I think I found the part of the video you were referring to. In this case, we could use a linear model to test the hypothesis that the difference in means is always 13 hours. To learn more about how to do this, see: ruclips.net/p/PLblh5JKOoLUIzaEkCLIUxQFjPIlapw8nU
Hypothesis testing is not commonly combined with machine learning, but you might use it if you wanted to be sure you selected the optimal algorithm. The hypothesis would be that both methods give the same results. If you rejected the hypothesis, you would know that one method was better than the other.
The video was explained really well, However why is the terminology used 'Fail to reject the null hypothesis', instead of just something like 'we accept the null hypothesis'?
We never "accept the null hypothesis" because there could because we can't differentiate between "there is no difference between drug A and drug B" and "someone made a huge mistake when handing out the drugs and gave everyone drug A". In both cases, the data will look similar.
I've got videos on confidence intervals here: ruclips.net/video/TqOeMYtOc1w/видео.html and ruclips.net/video/Xz0x-8-cgaQ/видео.html and area under the distribution here: ruclips.net/video/N4ZQQqyIf6k/видео.html
@@statquest Hey Josh, had a question As you mentioned about Rejecting the hypothesis and Fail to reject the hypothesis So does a hypothesis gets accepted also, if we get the same result as our hypothesis if we repeat the experiment? [Mesmerized to see that you still reply to the comments if anyone has a doubt :) ]
Hi Josh, This was awesome. But, can you please make some video regarding GCN(Graph Convolution Network) in future? It would be nice to get some explanation from your perspective.
Hi Josh,Thank you for the awesome video :) I'm slightly confused on how to frame null hypothesis for a given statement. For eg :A pizza place claims that their delivery time is 30 minutes or less than that on an avg. Now our null hypothesis will be "The delivery time is NOT less than or equal to 30 mins that is basically greater than 30 mins." Is this correct ?
Preliminary data is like a pilot study or just a small experiment to get a general sense of what is going on. It's not a full scale study or full scale experiment.
Is this a rule to say "no difference between things" when defining null hypothesis or can we also say "there is huge difference" as null hypothesis, if no then why?
When we say "no difference", we can put a specific number on that, 0, and the null hypothesis works as expected. BAM! In contrast, what specific number do you want to put on a "huge difference"? 12? 317? In order to do a test, you have to use a specific number. And how do you know 12 is better than 317? If you watch this video carefully, you'll see why you can't say "there is a huge difference" and use that as an effective null hypothesis.
NOTE: At 7:49 I meant to say Drug C instead of Drug A.
Support StatQuest by buying my book The StatQuest Illustrated Guide to Machine Learning or a Study Guide or Merch!!! statquest.org/statquest-store/
Hey josh i suggest you to write a book on everything you explained in your ML and prob&stats playlist. It will surely be a triple baaammm in the community. And also going to be best seller.
One day I'll do that.
@@statquest We are waiting! :)
@@GauravSharma-ui4yd Definitely, I was thinking the same, best seller without a doubt.
If only there was some nice feature to add quick, like, _annotations_ to videos to point out small stuff like that. Jeez, my hidden genius is scary =_= /s @RUclips
"I'm not going to name names", hilarious 🤣️
:)
@@statquest have you created a detailed video on inferential statistics ????? (•‿•)
@@omkargowda5764 I believe this video would quality as a detailed video on inferential statistics.
COVID 19
You know what I love about your channel Josh? You've made a real effort to get to the bottom of these concepts. It's beautiful to see such attention to detail and love for a subject.
Wow, thank you!
Videos on platforms that require subscription fees are not even a tiny bit to your content 😂👏Thank you so much for the work!!
Thank you very much! :)
really want to give you a big warm hug. You not only tought me these concept, now I am even confidence enough to explain these to someone else
BAM! That is awesome.
Josh doesn't accept hugs but BAMS!!
I literally watched your Statistics Fundamentals playlist in 2 days. It was too much of wisdom. I need to sit back, go through my notes again. Crazy crazy learning. Thank you so much for putting so many efforts that you made them look so effortless. Thank you, sir!
BAM! :)
My favorite hobby of the quarantine is to watch statquest and make notes. Very grateful to you Josh...
bam!
Josh's voice is so cathartic to listen to and there's a perfect balance of learning at a great pace with minor sarcasm/banter that makes these videos so easy to binge and learn from. Thank you sir.
Thank you so much! BAM! :)
And I'm extra impressed to see that you read and respond to comments on 3-year-old videos! That's a lost art in the RUclips-verse. Double BAM!
it just hit me that I laughed five times while learning statistics from this video. So far statistics lessons were about frowning and what not. BAM!!!
Bam! :)
This is the best statistics class I have ever been in, you break it down in a constructivist way, such that any rookie can understand
Thank you!
You explained better than my college lecturer
Thanks!
Will take my comps exam for grad school this coming saturday. Been using your videos to brush up on quanti methods. As a visual person, is so helpful. Thank you so much!
Wish me luck! :D
Best of luck!
Josh... and all of you friendly folks of the genetics department. Thank you guys SO MUCH for your great work. Thanks to you I learned so much in an intuitive way.
bam! :)
You're the Best.......It's like watching fun videos but in the end, we get amazing knowledge. Keep doing such amazing work.
Thank you so much 😀
Finally found the god of statistics
:)
This is so good! Thank you so much! I tried watching other videos about this but none could explain it well as you do nor keep my attention long enough.
Thank you!
i love this, i actually understand bless your heart
Thanks!
Thanks, now I mixed up all I have ever known about hypothesis!
Does that mean that this video helped you understand hypothesis testing, or did it just make things more confusing?
I'll be honest. I watch your videos waiting for the BAM and forget to focus on the concept! Then I've to go back all over again! 😂😂😂
BAM!
fantabulous class....a big salute
Thanks!
Literally I have an exam tomorrow. You’ve been a lifesaver!
Good luck! :)
Me seeing this comment a night before my exam: 👁👄👁 same bruh (ik it’s late reply but whatever)
@@-why6039 hahaha i get it :,)
@@aryananand5089 ikr! Statistics is 😭
@@-why6039 all the best! 😊✨
thank you so much for your videos
Thank you!
Thank you very much.
:)
your the best professor i ever had 👍
Thanks! 😃
Thanks Mr Josh for the wonder video.... You are taking lots of effects to make the video....... your clarity of explanation is really good .....
possible do it for anomaly detection in time series ........Thanks a lot for making more videos.
Thank you very much! :)
Jesus thank you man! I finally understand that null hypothesis it was tricky for me
Bam! :)
I'm supposed to be working, but I keep watching these
BAM! :)
You are such a cool teacher! Thank you
Thank you! 😃!
Very helpful video!
Thanks!
Thanks you for the video. It always refreshes my mind
Thank you! :)
I've learned something today.
Hooray! :)
@@statquest 000000⁰⁰⁰⁰⁰000⁰⁰⁰0⁰⁰0⁰0⁰0
Lovely BAM!!!!! from Bangladesh...
Thanks!
Using your page as a review for a job interview
Good luck! :)
Of course like! Awesome lectures! Explained very clearly and easily - the best I found. Unfortunately, I did not find anything 'clearly explained' at the channel about Markov models.
Markov Models are on the to-do list.
Great! And is a theory of random processes on the list?
these videos are amazing! you explained better than my statistics professor can. You earned a new subscriber! keep up the great work :D
Thanks, will do!
You know, I'm going to learn the chords of that cute intro! 😊
C, Aminor, G, C. BAM! :) (NOTE: They might sound a little different played on a normal guitar. The notes will be the same, but the order they are played might be different.)
thanks for this video
Thanks!
Thank you sir
Welcome!
I so wish I could meet you in person. You are just great!
Thanks!
I am your fan . Great help in machine learning
Thank you! :)
Hello Sir, thank you so much for explaining it so well. The examples given by you makes is so easy for us understand it. You are amazing! Thank you 😊
Thanks!
"I'm not going to name names" ooooh spicy time
:)
made it so easy to comprehend
Thanks!
Thank you! I'll be back.
Hooray! :)
i do learn a lot from you and enjoy much fun!!!Thank you Josh!!!!!!!!!!!!
Awesome, thank you!
Thanks Josh for the intuition & crystal clear explanation for Hypothesis Testing - one of the toughest topics to explain to someone. TRIPLE BAM !!! @statquest
Your explanation is better than any other resources FREE or PAID on the whole Internet. TRIPLE BAM !!! @statquest
Thanks!
Thank you!!
:)
Thank you
:)
This is REALLY helpful.
Thank you! :)
Your stat fundamental series realy helped me understand the basics of hypothesis testing, huge thanks for that! I searched for a video explaining the mathematical logic behind z and t test but found no one that realy explained the Details in an understandable way. Would you mabey consider making a video based on all the stuff you allready explained but connecting the dots and other neccecary details for z and t tests? Would be great if you'd think about it. (Unless you allready made one and I just didnt found it^^)
You know, one day I would love to do a proper video on t-tests. For now, however, the only thing I have is a video that shows you that the logic behind t-tests is the same exact logic behind linear regression, which is also the same exact logic behind ANOVA and a bunch of other tests (some of which don't even have names). For details, first learn about linear regression... ruclips.net/video/nk2CQITm_eo/видео.html and then learn how it naturally extends to t-tests... ruclips.net/video/NF5_btOaCig/видео.html
Great timing! Right after our discussion on NHST! :) ...another great video!
PS. You probably already caught this, but I believe there is typo at 7:50 ("drug A", but meaning drug C?)
Thanks for catching that. I added it to a pinned comment.
makes a lot of sense
bam! :)
Amazing videos! I just iked and subscribed double BAM!
Hooray! Thank you! :)
Excellent explanation
Glad you think so!
Hilarious! I've never lol'd on a stats video before
Bam! :)
Thanks Josh, you made my day! thanks
Hooray! :)
Thanks
Happy to help! :)
thank you!!!
:)
Josh- what about scientific hypotheses regarding forensic cases- where the null necessarily must point to 'not guilty' and therefore a fingerprint match null would be "different" and the alt "not different"!
I'm not 100% sure I understand your question, but you could imagine using DNA and saying "the null is that these two samples are the same" and then showing that there are enough differences to reject that hypothesis.
Thank you josh .... You are wonderfulllll😊
Thank you very much! :)
thanks !!
:)
Good one!!
Thanks!
Thanks for translating into Korean
Bam!
Triple Bam 💥 !!
Awesome Video for hypthesis testing
Thanks!
Thank you very much!
You're welcome!
Thank u
:)
You should hire a singer for the intro! ? 😂😂😂🎉
:)
trying to understand null hypothesis no one was explaining it as thoroughly as u thanks ! 🙏🏼🫶🏼
Happy to help!
I am trying to calculate the sample size to look up for five alleles frequencies in my population and after searching I found this equation: Sample Size = [z2 * p(1-p)] / e2 ,but someone just suggested that it should be calculated using the hardy equation!!! that's why I have asked if it can be used for sample size calculation!
I'm not familiar with that and did a quick search and wasn't able to find anything.
Hello! Do you think that I could get a basic notion of inferential statistics after watching the whole "Statistics Fundamentals" videos? If not, which ones do you recommend me to watch? :) BTW this videos changed my belief that I could actually understand something with numbers involved
Yes, I think you could get a good sense of what inferential statistics is and what you can do with it from my videos. Glad my videos are helpful!
Shoutout to josh. Homie got me an A
BAM!!! Congratulations! :)
Tack!
Thank you very much for supporting StatQuest!!! It means a lot to me that you care enough to contribute.
This is the first time I've heard of the null hypothesis defined as there being no difference between observations in test results. I always interpreted the null as the observations being due to randomness instead of being due to some causal relationship. This definition would be more in line with the p-value being the probability that the observations actually are random.
Of course, I could just be super confused :P
Another way to think of the null hypothesis is that all of the observations were randomly selected the same distribution (instead of two, distinct distributions). If all the observations came from the same distribution, then, in theory, there should be no, or relatively little, difference bean their estimated means.
@@statquest 🤯
How many experiments should there be to be confident in rejecting/failing to reject the null hypothesis?
There are ways to determine the number of experiments needed and I describe them in these videos: ruclips.net/video/Rsc5znwR5FA/видео.html and ruclips.net/video/VX_M3tIyiYk/видео.html
Great
Thanks!
Error at 6:16, arrow should point towards green balls.
No, the video is correct. At 6:16 we are comparing the red dots in the middle graph to the red dots in the graph on the right (at 6:18 )
Great video however, I have a question.
Throughtout the video, you used the phrase "random differences" for factors like Better diet, more exercise, (or may be some genetic mutation) which affect recovery time but are these factors really random ?
These factors may be unknown to the experimenter but is it a good idea to consider them random ?
I mean the phrase "Drug A may perform better than Drug B due to chance" doesn't make sense considering there would be deterministic reasons with mechanical explanations which would explain the action of the drugs
I agree, statisticians use words like "random" and "error" in strange ways. However, in the language of statistics, "random" means "things we didn't factor into the model". Unfortunately, that's just the way they use the term and there's not much we can do about it.
@@statquest thank you for the answer
Woh man, great content.
Thanks!
Great video, thanks! I completely understand what you said about us only being able to reject or fail to reject the null hypothesis, but I have seen many published papers stating their main hypotheses to be "there is no difference between groups". They then state in the discussion "the results support our hypothesis" when the tests find no difference and claim that the intervention is equally helpful for different age groups, for example. Is this problematic? Should the peer review have said something? Should they have switched the null hypothesis and the alternative hypothesis somehow because the type I error becomes the type II error and vice versa? What would be the algorithm for doing such research when they want to prove that the intervention is as effective as the other one?
Without knowing all of the details, I would say it is very odd to say that the main hypothesis is the null and that a large p-value supports that. The reason it's odd is that there could be a million reasons why the p-value was large, some of which have nothing to do with the treatment (for example, it could be that someone just wrote the data down incorrectly).
Hey josh... Hope you are doing awesome.... I am becoming better n better.... MEGA BAMMMM
Hooray!!! I'm doing well. I hope you are doing well in Chennai too.
@@statquest chennai wants to have duet with you... Eagerly waiting
This is awesome...can u please make video on Skewness and Kurtosis ... explaining it's concept and practical application.... Thanks in advance for reply
I'll keep it in mind.
Hi Mr Starmer, the lecture overall is incredible. Im just stuck at one point where I think the explanation flow is not as smooth as others. I know, BY COMMON SENSE, that the recovery time difference when giving drug C and D the 2nd and 3rd time is different, but not enough to reject hypothesis 1. Can u give me perhaps a more detailed and scientific explanation about why we simply are not certain to claim hypothesis 1 is incorrect, even though it is different then the following 2nd and 3rd time? Like how do we know if it is very likely that the hypothesis can be rejected?
What time point, minutes and seconds, are you referring to?
OK. I think I found the part of the video you were referring to. In this case, we could use a linear model to test the hypothesis that the difference in means is always 13 hours. To learn more about how to do this, see: ruclips.net/p/PLblh5JKOoLUIzaEkCLIUxQFjPIlapw8nU
@@statquest thanks a lot, highly appreciate it
You're so good!
Thank you! :)
peanut butter and jam !!
:)
When do we perform hypothesis test during machine learning project
Hypothesis testing is not commonly combined with machine learning, but you might use it if you wanted to be sure you selected the optimal algorithm. The hypothesis would be that both methods give the same results. If you rejected the hypothesis, you would know that one method was better than the other.
"* Bam!!" - Josh Starmer
Thanks!
You are amazing
Thank you! :)
The video was explained really well, However why is the terminology used 'Fail to reject the null hypothesis', instead of just something like 'we accept the null hypothesis'?
We never "accept the null hypothesis" because there could because we can't differentiate between "there is no difference between drug A and drug B" and "someone made a huge mistake when handing out the drugs and gave everyone drug A". In both cases, the data will look similar.
Bruh who is out here disliking this video?
Great question! :)
I'm just confused about how confidence intervals and area under the distribution come into play. Can you make a video combining these topics?
I've got videos on confidence intervals here: ruclips.net/video/TqOeMYtOc1w/видео.html and ruclips.net/video/Xz0x-8-cgaQ/видео.html and area under the distribution here: ruclips.net/video/N4ZQQqyIf6k/видео.html
I'm trying to determine what causes the bams to be smaller or larger
See: ruclips.net/video/i4iUvjsGCMc/видео.html
53 seconds ago and 9 likes already. People r fast.
Bam! :)
Hi, what is your hypothesis is that there is no difference between the two groups? I.e comparing anxiety in boys vs girls?
Generally speaking, the null hypothesis is that there is no difference and then we see if the data suggests that we reject that hypothesis.
@@statquest Hey Josh, had a question
As you mentioned about Rejecting the hypothesis and Fail to reject the hypothesis
So does a hypothesis gets accepted also, if we get the same result as our hypothesis if we repeat the experiment?
[Mesmerized to see that you still reply to the comments if anyone has a doubt :) ]
@@Actuay Technically you can only fail to reject the hypothesis.
Hi Josh,
This was awesome.
But, can you please make some video regarding GCN(Graph Convolution Network) in future?
It would be nice to get some explanation from your perspective.
I'll keep that in mind.
Small bam, bam, double bam, triple bam, may be the best video to describe how bam circle works on this channal😂🙏
BAM! :)
Am I the only one binging this stats playlist like a fiend?
bam! :)
Hi Josh,Thank you for the awesome video :)
I'm slightly confused on how to frame null hypothesis for a given statement.
For eg :A pizza place claims that their delivery time is 30 minutes or less than that on an avg.
Now our null hypothesis will be "The delivery time is NOT less than or equal to 30 mins that is basically greater than 30 mins."
Is this correct ?
Yes.
I'm always confuse with preliminary data..what it is stand for?
Preliminary data is like a pilot study or just a small experiment to get a general sense of what is going on. It's not a full scale study or full scale experiment.
Watching this during the pandemic, this feels surreal. Are you one of the Simpsons?
:)
Is this a rule to say "no difference between things" when defining null hypothesis or can we also say "there is huge difference" as null hypothesis, if no then why?
When we say "no difference", we can put a specific number on that, 0, and the null hypothesis works as expected. BAM! In contrast, what specific number do you want to put on a "huge difference"? 12? 317? In order to do a test, you have to use a specific number. And how do you know 12 is better than 317? If you watch this video carefully, you'll see why you can't say "there is a huge difference" and use that as an effective null hypothesis.
@@statquest Thanks for the clarification.