The F statistic - an introduction
HTML-код
- Опубликовано: 22 авг 2024
- This video provides an introduction to the F test of multiple regression coefficients, explaining the motivation behind the test. Check out ben-lambert.co... for course materials, and information regarding updates on each of the courses. Quite excitingly (for me at least), I am about to publish a whole series of new videos on Bayesian statistics on youtube. See here for information: ben-lambert.co... Accompanying this series, there will be a book: www.amazon.co....
My god, this playlist is saving my life.
This guy is god-tier at explaining statistics, I mean, there is no other online comparison that can be made.
YEaaaah same, I literally love him for making these.
can't sayf or the rest statistics channels on youtube. but this channel sure as hell teaches better than my professors whom monthly salary can send a kid's uni tuition.
damn!
Hahajaja
you have been incredibly lucid and I am recommending this channel to everyone who is struggling in econometrics! Thank you Ben!
Big thanks Ben, reviewing your videos before my intro to econometrics class really helps me clear things up. Thanks again.
This explanation is very very clear. I now have an idea of what the F test in linear regression is. Thank you so much sir
This is a really clear explanation of what an F test is. Great job and thank you so much for posting this on RUclips!
Different notations from what I'm used to ('RSS' and 'M' for no. of restraints) but still helped understand it all again perfectly, cheers.
Wish I saw your videos earlier... You are a life saver. A huge thanks from China.
If the value of F-statistics is larger, there is a higher probability of rejecting the null hypothesis. That means there is a higher possibility of a significant correlation between Y (result e.g. Sales) and X (variable on which dependency is being studied e.g Advertisement on TV)
its 2021 and this video made me cry tears of joy, why are American profs so snobby that they cant explain things as easily as you?
" it's unlikely that we're gonna fail to reject this hypothesis " man i hate this subject so much
Great video!
Your video helped me a lot in understanding more about the f-test. You explained everything fairly fast, but nevertheless I was able to follow and understand. Thank you very much!
I think this is a wonderful explanation for the conceptual idea of F or F-test!!! Thank you a lot
honestly man's a better professor than the one i pay for at uni
nice explanation. i'm liking your channel immediately.
Great video that supplemented my stats course.
Thankyou so much this explanation is clear and concise.
This makes so much sense now
Wow so clear, incredible...
Excellent video
Hi ben, love your intuitive approach!
Keep them coming ;)
Thanks for the video
Thank so much for this
Got yourself a new like here, great job
I can not believe we can access this video for free.
very good sir thank you its really helpful
Ben, Thanks for an EXCELLENT series. On the F-Test I see a seemingly different interpretation in some videos talking about ANOVA and a ratio SSW/SSB (variance between and within group). I will be highly obliged if you can draw a similarity / contrast between your interpretation of F test and SSW/SSB (ANOVA: Omnibus Test) interpretation or point me to some online material. Best regards
Regression and ANOVA are identical, the interpretation is the same effectively
Hi there ben my undergraduate lecturer has formed the F-stat by dividing the numerator by the number of restrictions and the denominator by the number of samples minus the number of parameters , hope you can explain why she has done this?
PLEASE BE AWARE THAT IT IS NORMAL TO USE DIFFERENT NOTATIONS AS LONG AS YOU STAY CONSISTENTLY.
shouldn't r-squared of unrestricted model be larger instead, you have more regressors, thus should explain more variation of y than restricted model?
Other than it possibly taking a longer time, is there a reason why it's not preferable to do multiple t-tests rather than an f-test -> maybe because with multiple t-tests you can at least figure out which exact coefficient is a significant value?
isn't it n-k-1 instead of n-p-1 where k is number of regressors in the unrestricted model?
In my opinion, thats the exact same thing with the video! In the video, p=# of regressors. However, the df of (SSRr-SSRur) is not the # of regressors. It is # of restrictions(for example when Ho:b3=b4=0, # of restrictions equals 2)
I understand, thank you
How do we know that the test statistic follows an f distribution?
Brilliant !
Is this useful for frm exam
Amazing
Why SSR(restricted model)>SSR(unrestricted Model)
what happens after you calculate the F-test???
We got a high T stat teacher over here
i have no image in my head for restricted and unrestricted. visualizations for all this would have been helpful...
why do you take the difference of the two SSR and not just the ratio of the two? the ratio of two chi-square already follow an f distribution
like (SSR1/n1)/(SSR2/n2)
Should be SSR_R >= SSR_UR and not just greater than if im not mistaken?
you are right
Thank you very much for the Video! Lets say i dont want to use the F-Test. Can i also use the T-Test for each parameter? (I know it wouldnt make sense time wise but can i understand the F-Test as a „shortcut“-solution? Like instead of doing 3+3+3+3 i can also do 4x3?)
Can anyone explain the difference between a restricted and unrestricted model?
In a restricted model, all beta values are considered to be equal to zero such that the regression line is equal to the regression constant. The restricted model has the largest sum of squared residuals. In the unrestricted model, the sum of squared residuals is at minimum (line of best fit)
why do you go into a low voice when saying important stuff?
emphasis
to hide you the truth!!!
I ran a regression and got an F value of 1655, what does that mean?
Reject it, and if it asks again, keep rejecting. Actually, get a restraining order against that hypothesis cause it's a serial fraudster
Hi Ben what is the intuition behind dividing SSR by degrees of freedom (in UR)? (is that your are dividing all the unexplained things by the variables that are not regressors?
Mitch Amp The idea is that you want a gauge of how well the unrestricted model explains the data. How you evaluate this depends on the number of parameters, and the amount of data you have; in other words the degrees of freedom of your model.
would you say the degrees of freedom is all the data spots that you are not using as an explanatory variable (and therefore not using to explain Y directly)?
n-k (k regressors)?
Mitch Amp Not sure I would put it exactly like that. The number of parameters estimated is k. This leaves n-k data points free to vary essentially. See this video: ruclips.net/video/-4aiKmPC994/видео.html - hope that helps!
degrees of freedom as n-k(parameters estimated)
how come b1 and b2 are not free to vary because depending the data points you use won't they vary aswell?
what if you're just testing Ho: B1=B2 ?
It's about re-specifying the model so you have a coefficient you can test. You could re-specify your model to take w1 = x1 - x2. Then your overall model would be: y = a + D1w1 + B3x3+...+Bpxp and you would test a Ho: D1=0 (since if D1=0 then B1-B2 = 0 and so B1=B2) with a t test. If you just have one restriction in the null hypothesis, you can use a t test. for multiple restrictions you must use the F test as always.
"the idear is?"
Jesus christ why does this have to be so fucking complicated.
🍧
Wow, are we still telling students to look things up in tables in the 21st century?