Masterful teaching, like your other videos. Deeply appreciated. To make things even better, it would be great if you added in the video's description a link (or links) to other videos to which you make references during a session.
I really thought there was no hope for me to be able to learn this stuff, but wow, I understand it perfectly now. Such a good video...like seriously man.
You are the best:))))) Thank you so much for the videos......I felt stupid in class but now feel so confident to even take an exam after your explanations.
Nice video, it is a really good complement with all the reading I have done in order to summarize the topic. Also nice examples, it is encouraging to learn not only the algorithm perse but also other topics
your Method of teaching and explanation is fantastic , but you use hard examples in explaining just like in 0:39 it took me a while to understand the question just use simple examples and you will be the best ever in statistics
Thanks for the feedback and compliments! Unlike most others on TH-cam, I use real data and real examples. I'm not a big fan of the made up "hours of studying vs mark on exam" BS that many others do. Trust me, that would make things much easier on my end, but I don't think it provides the value that real-life situations do. Most of my examples are still quite straightforward, but sometimes there can be complications (as there often are in real life). I don't think the example in this video is that far out of line, as in the end the variables are density and hardness. Cheers.
im confused about one thing. At 6:15, is that line based on the data from the whole population or just from a sample? The dots represent observed data, while the line predicts what the theoretical observed value should be, is that right? The line isnt the estimated regression line but rather the line formed using info from all data points in the entire population?
Yes, that's correct. In that plot, I'm pretending that we know the true values of beta_0 and beta_1 (the parameter values), rather than their estimated values based on sample data. In practice we never know the values of beta_0 and beta_1, but I'm pretending that we do in an effort to illustrate the theoretical model.
Y (without the hat) is a random variable, and when it takes on a value this will be an *observed* value of Y. Since the observed values of Y vary about the line (they don't fall precisely on the line), we need a random error component to represent that. Epsilon represents that random error component. The *predicted* values of Y fall precisely on the least squares regression line (the line represents the predicted values of Y), so we don't need a random error component.
This was superb video. But I am now confused between linear and non-linear regression. Suppose that I want to fit y=a+bsin(x)+csin(2x) when I have y=[0.39 3.13 2.87 4.62 2.27 0.34 1.53 1.61 0.75 -1.32 -0.032 -0.033 0.54 0.78 4.78 2.91 1.82 2.23 1.89 1.5] and x=[0 0.43 0.96 1.35 1.89 2.39 2.99 3.39 3.43 4.43 4.93 5.44 5.99 6.49 6.98 7.46 7.91 8.41 8.98 9.48]. Is this linear or non-linear regression problem? And why? Here, I am confused because of presence of sin which makes y to increase and decrease. The simple plot of (x vs y) is not very pleasing enough.
In that specific formula, you cannot simply substitute X bar and Y bar in. It would change the meaning, and would work with the assumed model. The *estimated* least squares regression line always passes through the point (X bar, Y bar), but that doesn't imply we can substitute them into the 6:21 formula.
it can be a bit misleading the way you introduce the assumption of a linear relationship.. linear regression refers to linear relationship in the parameters and Y. that's why fitting a quadratic polynomial for example would still be linear regression even though it isn't linear in the relationship between Y and X. otherwise, good video.
Yes, that is correct, and I agree with you in spirit. But in the context of introducing *simple* linear regression, I think discussing the notion that linear means linear in the parameters is unnecessarily complicating the issue. When I discuss multiple regression scenarios, and include polynomial models, transformations, etc., I spend a little more time on what the "linear" in linear model means. Cheers.
@@kellynaz9256 if you referred to me as a male I wouldn’t get offended because I am a male lol. What’s offensive about calling you what you are? You are a female aren’t you ?
Another video that fails to explain that within this field the variable names can differ from person to person, making EXTREMELY CONFUSING for newbies.
jbstatistics Certainly. Density of austrailian timber, Janka hardness, empathic concern scale. These are not examples everyone can relate to; it is needlessly complicated, especially for those who have come here wishing for it to be simplified. I never had a problem distinguishing which variable is x and which is y, but i struggled with your scenario. Why not just do advertising expenditure and sales? The rest of it was really well done.
Mujibul123 agree, the only part of the video that causes unnecessary complication. Way too esoteric, who can relate to the Empathic concern scale or janka hardness??
Bro you could just google those, those examples if u understand are soo made for regression!(At first even i felt the same just google, you'll understand)
The problem is that the example is so weird, that it can be distracting. Maybe you were trying to be creative, but it does make it a bit harder to follow. The rest was great, but I would consider using mundane variables such as exercise and weight.
It's shocking how simple these concept become when taught by someone engaging who cares and understands it well. Thank you!
WORD!
1 hour lecture summarized in 8 mins thnx pal
+okay You're welcome.
These videos, helped me ace STAT 2040 last semester, and are now helping with my new job. Thanks Jeremy Balka for being the best prof
8 years after and this video is still relevant and useful. Thank you so much for this!
You're welcome. I'm glad to be of help. Thanks for the feedback!
WOW! First, your up-beat tone makes learning this not so daunting. Second, I am finally understanding these concepts!!
Masterful teaching, like your other videos. Deeply appreciated.
To make things even better, it would be great if you added in the video's description a link (or links) to other videos to which you make references during a session.
I really thought there was no hope for me to be able to learn this stuff, but wow, I understand it perfectly now. Such a good video...like seriously man.
Thanks Aaron! I'm glad you found this helpful.
jbstatistics Sure! Where is the next video that talks about the "Least Squares" Method?
Aaron Jacobs Oh, nm. I think I found it. It's probably the one called "Simple Linear Regression: The Least Squares Regression Line" lol
Yes, that's the one. I'm glad you found it!
Thanks! I'm glad you find them helpful.
Thanks! I'm glad to be of help!
Where is another video?
Thanks! You are very clear and well paced, which made it easy to follow and understand
You are the best:))))) Thank you so much for the videos......I felt stupid in class but now feel so confident to even take an exam after your explanations.
I just discovered your videos. They are succinct and insightful. Thank you!
You are very welcome!
You are the best! Thanks from brazil
You are very welcome, and thanks for the compliment!
Hate from Canada
Thank you so much for these excellent lectures.
Didnt ask
Once again, your videos are a great help. You are great.
Nice video, it is a really good complement with all the reading I have done in order to summarize the topic. Also nice examples, it is encouraging to learn not only the algorithm perse but also other topics
Thanks for the compliment! I'm glad to be of help.
Concise and informative. Very good work!
Thanks!
You are the Sal Khan of Statistics!
your Method of teaching and explanation is fantastic , but you use hard examples in explaining just like in 0:39
it took me a while to understand the question
just use simple examples and you will be the best ever in statistics
Thanks for the feedback and compliments! Unlike most others on TH-cam, I use real data and real examples. I'm not a big fan of the made up "hours of studying vs mark on exam" BS that many others do. Trust me, that would make things much easier on my end, but I don't think it provides the value that real-life situations do. Most of my examples are still quite straightforward, but sometimes there can be complications (as there often are in real life). I don't think the example in this video is that far out of line, as in the end the variables are density and hardness. Cheers.
Great video! Thanks from Argentina
Incredibly well explained thanks!
I'm glad to be of help. Thanks for the compliment!
This will help with my A.I project. Thanks
you are the best keep on doing this i always understand from your video
im confused about one thing. At 6:15, is that line based on the data from the whole population or just from a sample? The dots represent observed data, while the line predicts what the theoretical observed value should be, is that right? The line isnt the estimated regression line but rather the line formed using info from all data points in the entire population?
Yes, that's correct. In that plot, I'm pretending that we know the true values of beta_0 and beta_1 (the parameter values), rather than their estimated values based on sample data. In practice we never know the values of beta_0 and beta_1, but I'm pretending that we do in an effort to illustrate the theoretical model.
+jbstatistics thanks!
Your videos are in the 'What to study' section of my university course ;P
Then you've got a smart professor :)
This is an awesome lecture
thank you! this video is really helpful...much easier to understand for beginners
You are very welcome!
I heard the word Jenka hardness I am out.
hahaha so true
thank you, i need this so badly :)
Great as usual. Thank you :))
You are very welcome. Thanks for the compliment!
at 7:25 can you explain the reason of not using _epsilon_..I still didn't get it
Y (without the hat) is a random variable, and when it takes on a value this will be an *observed* value of Y. Since the observed values of Y vary about the line (they don't fall precisely on the line), we need a random error component to represent that. Epsilon represents that random error component. The *predicted* values of Y fall precisely on the least squares regression line (the line represents the predicted values of Y), so we don't need a random error component.
Okayy!!. Got it..
Thank you
This is so helpful!!
Great stats videos!! do you also have videos to multiple regression ? couldn't find any on your page...
This was superb video.
But I am now confused between linear and non-linear regression.
Suppose that I want to fit y=a+bsin(x)+csin(2x) when I have y=[0.39 3.13 2.87 4.62 2.27 0.34 1.53 1.61 0.75 -1.32 -0.032 -0.033 0.54 0.78 4.78 2.91 1.82 2.23 1.89 1.5] and x=[0 0.43 0.96 1.35 1.89 2.39 2.99 3.39 3.43 4.43 4.93 5.44 5.99 6.49 6.98 7.46 7.91 8.41 8.98 9.48].
Is this linear or non-linear regression problem? And why? Here, I am confused because of presence of sin which makes y to increase and decrease. The simple plot of (x vs y) is not very pleasing enough.
Very good! Thanks!
You are very welcome!
Was doing the Andrew Ng's course and damn I couldn't understand this, but thanks to you :)
I'm glad to be of help!
6:21 - Would X be the same as X-bar in this case? (and the same for Y?)
In that specific formula, you cannot simply substitute X bar and Y bar in. It would change the meaning, and would work with the assumed model. The *estimated* least squares regression line always passes through the point (X bar, Y bar), but that doesn't imply we can substitute them into the 6:21 formula.
Oh I think I understand now... thanks Mr S.
That was pretty neat!
Thanks for the compliment!
Dude, thank you! Subbed.
this video is excellent
Thanks for the compliment!
great vid
What is intercept. I could notunderstand wt is it. Also how t odeop the trendline and wt si the starting and ending point of it?
Best explanation 100//
Thanks!
Finally I understand what those ß are...
I'm glad to be of help!
How do you estimate the significance of the betas?
thanks for you
Great videos =)
Thanks!
Thanks Sir!!!!
You are very welcome!
nice video
do u have something on correlation?
Is it another name for Bivariate Regression?
thanks!
You are very welcome!
it can be a bit misleading the way you introduce the assumption of a linear relationship.. linear regression refers to linear relationship in the parameters and Y. that's why fitting a quadratic polynomial for example would still be linear regression even though it isn't linear in the relationship between Y and X. otherwise, good video.
Yes, that is correct, and I agree with you in spirit. But in the context of introducing *simple* linear regression, I think discussing the notion that linear means linear in the parameters is unnecessarily complicating the issue. When I discuss multiple regression scenarios, and include polynomial models, transformations, etc., I spend a little more time on what the "linear" in linear model means. Cheers.
Janka, Janka, Janka, god I am fucked for my final exam...
Who has an inferential stats exam tomorrow?
🙋🏻🙋🏻🙋🏻🙋🏻🙋🏻🙋🏻
I hope your exam went well!
>Not using the metric system
"females" lmao i love how you said that like we are alien creatures, good video though
Stop getting offended about everything. Not everything is a personal attack, no one cares about you
@@aliaziz1145 and maybe you should refer to women as women not females ya knob. Dont get ur balls in a bunch
@@kellynaz9256 if you referred to me as a male I wouldn’t get offended because I am a male lol. What’s offensive about calling you what you are? You are a female aren’t you ?
I have my exam in 4 min lol
Another video that fails to explain that within this field the variable names can differ from person to person, making EXTREMELY CONFUSING for newbies.
Though the examples were tough but it's easy to understand.
aduh
Very confusing lol
Kindergarten mathematics for economists. No wonder the world looks like it does...
These are the worst examples, so badd!!
Care to explain why you feel that way?
jbstatistics Certainly.
Density of austrailian timber, Janka hardness, empathic concern scale. These are not examples everyone can relate to; it is needlessly complicated, especially for those who have come here wishing for it to be simplified. I never had a problem distinguishing which variable is x and which is y, but i struggled with your scenario. Why not just do advertising expenditure and sales? The rest of it was really well done.
Mujibul123 agree, the only part of the video that causes unnecessary complication. Way too esoteric, who can relate to the Empathic concern scale or janka hardness??
Bro you could just google those, those examples if u understand are soo made for regression!(At first even i felt the same just google, you'll understand)
The problem is that the example is so weird, that it can be distracting. Maybe you were trying to be creative, but it does make it a bit harder to follow. The rest was great, but I would consider using mundane variables such as exercise and weight.
This is horribly difficult to understand wtf
thanks for you
You are very welcome!
thanks for you
You are very welcome!
thanks for you
You are very welcome!