Interpreting Linear Regression Results
ฝัง
- เผยแพร่เมื่อ 26 ก.ย. 2024
- This video describes how to interpret the major results of a linear regression...
...so I just noticed that this video took off. Thank y'all. You are most kind. Yes, I made some mistakes. you are correct...the Y is the DEPENDENT variable and the X is the INDEPENDENT variable. When I recorded this, it was for my undergrad students and it was off the cuff. I appreciate that it has helped you get through your own stats class. This video is designed for an undergrad stats student that will probably not go on to do major research so, clearly, there are plenty of details and stats stuff that I didn't do.
The link to the data is: www.kaggle.com...
I love this dataset because I usually finish the lecture with a tableau demonstration and most everyone realizes why the prices of these homes are so high.....
I FINALLY UNDERSTOOD THIS REGRESSION TABLE, GOT A 100 ON MY STATS TEST BC OF U YOURE MY HEROOOOO THANK YOU VERY MUCH DR
I am humbled by your comments and encouraged by your success. Thank you and congratulations to you.
Best video I've seen on this topic. Thanks!
Love how you broke it down in to answering the 4 core questions and how to answer them
Thanks Sergio, I studied this years ago during my Degree but it got rusty for not using it. Now with your tutorial I remember it even better than then and I think this is one of the most detailed tutorials here! Most of them don't talk about the Alfa coeficient or not going on depth like you did. Thanks again!
thanks for the comment. I appreciate it.
I was so overwhelmed by the subject but you made it so easy to understand! Thanks Dr. Garcia~
Glad to hear it!
This is the best video on explaining significance. Learned so much from this!
Wow this helped me so much with my data analytics major project. Thank!
you taught in this Vedio more than I learned in my whole semester. Thank you very much for upload this Vedio
Hi, I like the way you explain statistics. I really appreciate the colors you use; I can focus better with no need to stop the video for a long time. Clear voice and speed just perfect. I'm an education psychologist graduate student; thinking on going further, probably statistics. Thanks for sharing.
Thank you so much, really helpful video for my dissertation project I'm working on at the moment. Only 3 weeks left 😢but this will definitely get me through it!
Good Luck! It'll be over before you know it!
this was an amzing video saved a student today thankyou
Thank Sergio for this detailed explanation for the linear regression, very useful when i need to explain the result!
good to hear. thank you
Excellent and straight to the point. THANK YOU!
Dr. Garcia is a literal god at statistics
He dident interpret the coefficients.
Superb!! Dr. Garcia guides how to pick up the right data points from the noise and how the relevant data points support answering the problem statement, thank you !!
This is soooo clear!!! thank you so much.
came for the content, stayed for the *SSSSSSSSSSSSSSSSIP* at 6:54. nice video
my students make fun of my coffee drinking all the time. lol.
I finally understand i have been struggling to understand this topic
Excellent explanation. Thank you.
Thank you so much! Great and concise explanation paired with easy-to-understand language!
you really helped me as i was doing my final year research project,cant thank you enough
Hello Sergio, this is an excellent walk-through; thank you for presenting this, and thank you for your dataset reference. I appreciate your clarification on your presentation focusing on a "single independent variable linear regression."
Thank you sir! This is such a great explanation, I was working on a project at work and you helped me a great deal!
Your video is very clear and easily understand the linear regression results. Thanks !
wonderful explanation, bravo
Super Explanation... Amazing
Thank you! Well Explained!
This is supper explanation. Thank you very much, it enhances my class work and add me confidence.
Confidence is key!
thank you for the explaining, it's clear a lot, but i have question what does mean by a good variable
It's a good video, it's ctrl+shift+down is a life savior
Truly amazing video!
Thank you Dr. Garcia, appreciate it.
Excellent video!
Very helpful!
Thanks for the video
my joy.
Thank you
Being a simple linear regression, the p value will be the same as the sig F. They are both testing the same null hypothesis (coeff = 0). I don't see how the answer different questions here.
no argument. but we're also setting up the future transitions into multiple regressions down the road.
Thank you very much for a wonderful explanation!
Glad it helped!
Thanks prof
Thank you so much for taking time to explain regression! This vid put my mind on rest! :-)
My joy. Thanks.
Thanks, very clear explanation
great
X is independent and Y is dependent variable
yes....*embarrased*
JazaKAllah ❤️very helpful
I thought the coefficient have to be between 1 and 0, my results are pretty similar to yours with high coefficients and I was confused?
There are other methodologies where that is true, like a logistic regression. You may also be confusing coefficients with traditional probabilities......(emoji with guy shrugging his shoulders here).
Such a perfect video ,thank you so much.
Sir thank you for your clear demonstration just got a question, if I am right the video is doing a single-tail test, if I am doing a two-tail test is it the same way to accept or reject the H0 by comparing the p-value of the coefficient with the significance level(i.e. 5%)?
The Ftest by definition is a one tail test. You can't do a two tail test in a regression. Technically, it is the same process when testing the pvalue, but it is a different distribution.
@@sergiogarciaphd7142 Thank you for the reply sir but I still wonder why a two-tail test cannot be carried out in a regression and also what distribution are you referring to? I apologise if the question is unnecessary, but I would greatly appreciate it if I can hear further explanation from you.
Thank you soooooo much.This is a great explanation!!!!!
your happiness is my happiness.
Nice
Fantastically explained ! Thank you very much for having taken your time to help us
Thank you. Good luck in your course.
The ultimate goal of running a regression is to be able to predict/estimate the dependent variable. Right? How would you predict if the intercept is negative when you know that neither dependent or independent variables can be negative? Can we force it to zero intercept? If so, how?
THAAAAAANK YOU..JUST ABOUT TO DO MY EXAMS
You are amazing! Thank you.
thank you!!
Doesn’t X is independent variable and Y is dependent variable?
yes..yes it does.....(#facepalm)
Yes
th-cam.com/video/PaYqpOc-YNI/w-d-xo.html
i want to learn how you connected a pen to you computer. what pen do you use? This is amazing how it helps in explaining and understanding
I have a Wacom tablet, which comes with a pen, and a screen capture software tool. easy cheesy.
Wow!!! Loved it!! 👍🏻👍🏻👌
thank you a lot I am 2 hour a way from my exam and I was winder when F-significance rejects the null hypotheses 🙏
Excellent! Thank you so much!
Glad you liked it!
How cool is that?!
It's the COOLEST!!
Can i use regression analysis for data collected at irregular intervals?
maybe...but the words irregular intervals suggest you are running a time series analysis and that requires a different set of assumptions. The results of a simple linear regression may be misleading for that type of analysis, which is not to say that it sounds like you may have a lack of data as well.
Thankyou for sharing!
thank u Dr. Sergio.
Thank you.
THANK YOU SO MUCH ,BUT I GET LOST ON SIGN WHY WOULD YOU SAY TS SMALL YET WE HAVE 2.33 MORE BIG THAN 0.01
You may be confusing the actual statistic with the P-value here. I am assuming that you are looking a classic back of the textbook table here. Consider what it is that you are getting out of the table and do not confuse the tstat (on the axis) with the pvalue (area under the curve).
If x variables are inputs, shouldn't those be independent variables? Since the output (y) depends on x variables?
yes. you are correct. That is my mistake and it keeps me humble as more and more people watch (and correct) the video.
@sergiogarciaphd7142 Thank you for confirming. Your videos helped me pass my finals. I didn't forget this one since I had to dig. Worth it.
Great videos!
Thank you for the great video..can you explain the analysis of standard error?
Hii, I have done my data analysis and the p value for the model as a whole is not significant, but one of the independent variable is significant. Can I still report it's results ?
This is so good
Thanks for the video! Can you please send the link to the dataset?
link is now included in the description! thanks.
the intercept's p-value is more than 0.1 How to interpret this
Hi, it is really very helpful, clearing many doubts in an easy way. the one thing which I would request you to help me understand is why you were referring P value as 0.0000, however it is 2.416 in the regression model. please help what is that I am missing here.
Can anyone help me explain these numbers? coefficient: .049652, _cons .5150066, t 6.12 t 68.62 p>t 0.00 0.00 number of obs 6.665 prob > F 0.0000, R-squared 0.0056, I am trying to find out whether socioeconomic status can influence whether students find it important to perform better than other students in tasks. I am really bad at understanding these numbers.
this is the most motherf*uckin' superb lecture I've ever attended! Thanks for it!
p.s. just create a full udemy course related with econometrics or data analysis, u'll earn thousands of dollar from it for sure!
Thank you so much! :)
Perfect.
Hi, thanks. I have a large dataset of 100,000+ points. How does this affect the linear regression analysis? I got a very large F, an extremely small p value, and a small R^2.
This can be tricky....you need to talk to your stats person directly. Sometimes when you have a very large number of observations, everything is significant. You'll want to talk to someone directly about your work and get better guidance.
Use of tstat?
Great video! 4:09 May I ask what drawing software are you using to annotate the Excel file? Thanks.
The new Excel allows for drawing right into the spreadsheet..the real trick is to get a screen that allows the use of a pen..like the Wacom tablets....
If my f = 0.013. Does this mean It doesn’t have significance?
Great video, I understand now. Only thing I don understand is the "alpha", where did that come from? The smaller value of alpha means more tighter requirement?
You get to pick the alpha. (Your professor/supervisor/client will do this in practical application.) but when you are the decision maker, you pick the alpha...consider manufacturing where we use the six sigma process for quality....the alpha is, for all practical purposes about 0.01....that becomes the industry standard and it is in essence picked for you....but you get to decide the alpha...and more to the point when you use the decimals in the way you see it in the video...you can easily test any alpha.
my questi
on is how is x axis dependent variable usually its the independent one
Yes. This has been addressed in comments many times. It was my mistake and you correct. The Y is Dependent and the X is Independent.
thank you so much
Sergio, thank you so much! This helped a lot as I am working with this now the first time in my bachelor thesis. I wonder were the alpha comes from? Did I miss anything? Thank you again :)
You get to decide the alpha. (When doing homework, the problem or your professor will decide the alpha.) In the real world, you get to pick. This is implicit in applications such as six-sigma quality stuff...think ISO9001...but in reality, and certainly in academia, we know and report the three big alphas..(0.10, 0.05, 0.01) all at the same time by using those astrixes (or whatever the plural for asterix is). We can tell by the result what the highest alpha is where the variable is still significant. ;)
Can we conduct logistic regression analysis using excel?
It was really helpful thank u so much
You bet. good luck.
Great Explanation! I watched all your videos w.r.t univariate & multiple linear regression on the same dataset and I too tried doing this. It works fine but i have a question(This question is w.r.t your other video on multiple linear regression on the same dataset) - Why have you removed waterfront, view and grade columns from the dataset? I understand those columns might not be required for the regression analysis but if I keep those columns while doing multiple linear regression, it seems like these 3 columns also impact price a lot. P value is 0 for view and grade as well which is less than 0.05 & 0.01.
Could you please explain this? if we assume these columns dont impact price value then their P value should be greater that 0.1 right?
Looking forward for your response.. Thanks!
I don't remember the results you are referencing off the top of my head, but there is a certain art to selecting variables of interest and measuring the impact. This is not usually a discussion for undergraduate courses but the impact of additional variables is a discussion for MULTIPLE linear regressions. That's a whole other ball game. This video is about interpreting the results of a simple single independent variable linear regression.
I thought "Y" was the dependent varibale?
It is. (facepalm emoji) that one flub will be with me for eternity...
have u performed multiple linear/nonlinear regression
What is alpha?
Why does rejecting the null mean it is a good equation please? Thanks
This has to do with the way you set up the hypotheses and is past the scope of this one video. That question comes before you do the math to understand how the scientific method works.
what if R square is 10% but alpha and Sign F are less than alpha ? Can we say still my data predicts good model?
one thing at a time. If your P value is LESS THAN alpha, then you can interpret your R-square value. If it is not, then what's the point of R square? There is a larger more academic argument here about whether or not the relationship still exists and why we need alpha anyway...yaddiyaddiyadda........, but for most students and direct applications, check for significance before you do R-square.
Which pen do you use
Hi, I have produce a regression table just like yours in the vid, but with a different data sets. May I know the reasons why my significance-F is zero?
The significance of this F is also zero. Excell simply used the (E-70) to detail the decimal numbers are so small they have 70 zeros in front of them. Your zero is probably the same....some super small number that the software shows zero when displaying a smaller (usually 3) decimal places.
13:07 you rejected null hypothesis but isnt the p-value (2.417) bigger than alpha 0.01? Where did you get 0.000? Im confused
The p-value is NOT 2.417. It is 2.417e-70. which is 0.0000000...70 zeros....0002417. It is super super small. Excel is giving you a result of precision that is unnecessary and alerting you to the actual value by using the 'e'. This is NOT euler's constant. This is saying X 10^-70
@@sergiogarciaphd7142 Thank you so much for the clarification! Have a great day!
What Excel add-in are you using for the "Draw" panel in the worksheet?
It's the basic Draw function. The real trick to have a pen screen like Wacom.
@Sergio Garcia, Phd: Thank you for the video, it was very helpful. I am currently writing my master thesis and I had cited this video, my supervisor asked me to cite the original literature instead of this video. Do you have any research paper of your own or a book where you have formulated these 5 questions that one should ask while interpreting the regression results? Thanks :)
I do not. Any undergraduate stats book will tell you the same, you just have to learn how to read it. Your stats prof knows this and wants you to use your expertise which also relies on the texts you have already been presented with. Use your stats book and go back and read it again. It’ll make more sense. I promise.
Which good stats book you recommend sir?
What if I have only 20 observation but my R-squared is 0.99367. Is it normal for a small data to get this R-squared?
so....it's nice to have more observations..but that Rsquare is really good....almost too good?
good day sir, I just wanted to ask if an independent variable is not significant or does not have an explanatory power to the model but when removing it lowers the adjusted r-square what does this imply? so far the reason that i know the reason is because the t-statistic is greater than one. With this information, what can we infer?
It implies exactly what you described in the question. This is a classic trade-off situation. You, as the researcher, need to decide what is important. This is only one of many issues where researchers can disagree on results given the same data and methodologies. Sorry dude. no easy answer here. This is a classic, "What do YOU think?" question.
Is it a problem if standard error is around 300 if yes what might be the reason for it? I have taken small sample, so do I need to improve the data?
Hello, did you get an answer for this? I have really good R squate values and significance value but my std error is way too high
Sir how can I calculate y=mx+b equation
Hi Sergio, I have a question that has always troubled me. If the regression is significant (fvalue
uh...that's interesting, but not exactly what significance means....when we talk about significance, we are discussing the probability that the relationship between the two variables can/will be replicable if we run the test again with another similar/random/representative sample...and therefore predictable moving into future and we can then make decisions. So we're really talking about the big picture relationship between the two variables. I suggest you look at the residuals (Expected Value minus Actual Value) to see differences between observations. This means that some observations will be closer to their expected value, but that does not mean that any one point is more/less significant than any other.
@@sergiogarcia9044 Thank you very much for your reply! Regression analysis is often used in agricultural studies to examine the influence of rate on an outcome, say wheat yield. So, if three fertilizer rates, low, medium, and high, and applied to plots of wheat and the regression analysis result is that there is a significant positive linear relationship between fertilizer rate and wheat yield, can I say that the low rate is statistically different than the medium or high rate without performing pairwise tests between the yields from each rate of fertilizer? Or, is regression only saying that the relationship is significant, so if fertilizer is increased, yield will increase within the range tested (no extrapolation)?
@@4mfenme not really. In that scenario, you have 3 data points which is woefully small for a regression. You may be better served doing a t-test for differences between two groups which would show if there is a difference between the two groups or an ANOVA, which could include more than two. That is a classic hypothesis test question and often used in stats Exams..probably more so for Ag specific majors, but the point is the same. You gotta make sure you are using the right tool for the right question.
@@sergiogarcia9044 Thank you, Sergio!