Statistics 101: Linear Regression, Understanding Model Error
In this Statistics 101 video, we learn about regression model error. To support the channel and signup for your FREE trial to The Great Courses Plus visit here: ow.ly/xVD030fiZ8S
Now, this video is the next in my series on simple linear regression and it is an important one. In this video, we're gonna talk about understanding model error. So understanding error in regression is absolutely fundamental and crucial to getting a really deep understanding of what's going on in regression, how we can build better models, and actually understand what the output is telling us when we run regression in Excel or JMP or R, whatever it is, understanding exactly what we're looking at and I'm gonna do that in as visual a manner as possible. There will be a little math in here but nothing difficult. So let's go ahead and get started.
So if you are new to Regression or are still trying to figure out exactly what it even IS...this video is for you. So sit back, relax, and let's go ahead and get to work.
My playlist table of contents, Video Companion Guide PDF documents, and file downloads can be found on my website: www.bcfoltz.com
#statistics #regression #machinelearning
The Great Courses Plus is currently available to watch through a web browser to almost anyone in the world and optimized for the US, UK, and Australian markets. The Great Courses Plus is currently working to both optimize the product globally and accept credit card payments globally.
Пікірлер: 65
I have watched all your playlist from 1 to this one and will finish the remaining. I have learned more, with great depth and understanding of fundamentals, in one month with your videos, than what my MBA program taught me about data science in 2 years.
@santisav2
Жыл бұрын
Same for me
"it's simple math, don't freak out" yeah i felt that
You have explained and illustrated one of the most important topics in statistics really well! It’s a great video and highly useful.. Thank you so much!
I'm a masters student in statistics that's taking a first semester of intro probability theory and linear regression. Honestly, this series is actually making everything click on the application. Direct proofs of definitions and the textbook wasn't doing it at all. Brings motivation to answer why we're proving what we're proving
Hands down the best statistics video I have seen on KZread!
I've never left a comment on KZread but I had to this time. I'm so in love with your channel. Good content amazingly explaned. Thanks!
You´re really the best, thanks a lot Brandom
These days, I don't usually open KZread. But if I do , it's for completing this playlist and the others you've uploaded. Thank you for saving a distraught student.
Lecture was great but one thing i didn't understand that we calculated variance of x and y (bill and tip) by dividing "n" in the denominator (and not n-1, since it is a sample), but while calculating MSE we are dividing it by (n-2) because we considered it as a sample.
The content is so neat, you make stat simple and easy. Thank you Brandon :)
Great explanations! thanks a bunch!!
it is very clear about regression basics.
Nicely Explained. Thank you Brandon!
Thank you for making good informativ videos on this topic. It's hard to come by.
the explanation is very detailed. I like it very much.
Ur the best dude
Merci Monsieur
Perfect!
thank you!!!!!!!!!!!
tnx
Your amazing...
Making this vid the same length along with the same title would be cool
I love it !! learning so much, just a bit confused on the excel, and how to do that , is there videos with more explanation on how to do data analytics for regression in excel?
Geat!!!
. Thank you.. .
I really miss the motivation you used to give at the start of every video. Please include that motivation in every video lecture.
Do you teach Data Mining as well?
nice video! I actually find the R-square (coefficient of determination from your other video ) =74.93% , where the correlation r= .866 is actually square root of "R-square". is that a coincidence ?! the correlation of simple linear regression is actually square root of SSR/SST!
@AlexAlex-pe7mn
5 жыл бұрын
it is true, not a coincidence
what does this significance value tell us? I mean the significant difference is between what? Also, what is adjusted R square? What is the difference between R-squared and adjusted R-squared
what statistics software you use to calculate ANOVA and model error and F etc?
how do you get the standardized coefficient beta
how do u get the F value at 20:55 ?
Thank you for great video series. Sorry, in 13:08 , maybe you mean SSE divided on the difference of sample size and DF? )
@BasuthkarKiran
Жыл бұрын
Yea here 'n' represent the sample size. And 2 is the dof
Can anyone clarify me this point? At 12:50 Brandon says "MSE is an estimate of sigma square, the variance of the error epsilon". But isn't sigma square usually used to represent the variance of the population data? Why it is used now to represent the variance of the error?
I'm confused. Residuals have always been explained to be the difference between observed value to the predicted value. Here you say it's the observed value to the mean. SST = SSR + SSE, in which the SSR is the one that looks at the squared sum of residuals.
13:01 how did we figure out degrees of freedom as 2? Also 14.10.shouldn't it be n-1?
@surjeetbasu797
3 жыл бұрын
One way Anova ;for MSE calculation ,Degree of freedom : N-C .In this case C=2 , N-2 is our Degree of freedom for calculation of MSE.
When can we see the videos on time series?
May I ask, is the RMSE same as Standard Error of the Estimate?
@BrandonFoltz
4 жыл бұрын
Almost certainly yes! :) Different software can name it differently but root mean square error and standard error are almost certainly referring to the same thing.
what is mean of response?
@13:08 you are mentioning degrees of freedom as 2.. should'nt it be 1?? The Anova Table @5:18 shows 1 as degree of freedom for the model and 4(n-p-1 = 6-2 ) as degree's of freedom for errors..
@yizhang6258
5 жыл бұрын
Hi, I have the same question...
@n9537
5 жыл бұрын
for Simple linear regression, the degrees of freedom for SSE is n-2 because there are 2 quantities estimated(the slope and the intercept) which limit the "freedom" of the data points(in this case the squares of the error terms). So MSE = SSE/n-2
09:25min: the numbers of the squared errors are not correctly calculated I guess. Could you please confirm.
@BrandonFoltz
4 жыл бұрын
Hello! They are correct. Since they add up to the correct SSE and I do those calculations in Excel later in the video and they add up to the correct SSE they therefore are correct. If there is a specific issue you are having let me know! :) Thanks for watching.
@bcc1432
4 жыл бұрын
Thank you for your reply and your videos, they are really good!!
is it correct if i say that standard error is identic to standard deviation?
17:00 why does the model have 1 degree of freedom?
Hey Brandon, In some lectures they calculate r_squares as , r_square = 1 - (SSR/SST) But, You say r_square = SSR/SST Does this both contradict ?
Sir, What does it mean by Sample data and Population data? Can you please clarify my doubt?
@HeathenChannel
5 жыл бұрын
Sample is a part of the population. For example, if our waiter served 20 tables that night, that would be our entire population. Here we are analyzing a sample of six tables.
Try playback on 2x normal speed
What is F and Significance F?? The videos are great but it all falls apart when you assume that knowledge. Are we supposed to have watched all previous 13 playlist in full?
Degrees of freedom is 4 not 2. Looks like the explanation for degrees of freedom needs correction.
You didn't explain, what's Adjusted R^2
@BrandonFoltz
2 жыл бұрын
I made an entire video on it in June 2021. Unfortunately I sometimes mention things that are present in output that I haven't gotten to yet.
@prateek2159
2 жыл бұрын
@@BrandonFoltz oh yes I found it...thanks ✌🏻
It's not difficult but it's kind of hard to see the big picture
Merci !
@BrandonFoltz
Жыл бұрын
I really appreciate it! 🙏