Likelihood ratio test - introduction

[@S4Bier]

Thank you so much! When it comes to statistics and I have a problem with a specific topic, the first thing I do is to seach on youtube for "Ben Lambert Topic X". There literally is nothing in the whole internet which explains it better than you do. Godfather of statistics!

[@richardbeck2048]

can't describe how grateful I am for these clear explanations!

[@SufferingAddict88]

I wonder... why can't I ever have professors that can explain things well? It always the same. Your in the lecture, you don't really get what the professor is talking about. You go on the internet. Bam, someone explained is so well any fool could understand it.

[@lastua8562]

Not just any fool.

[@stinksmcgee]

XDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDD

[@JD-jl4yy]

Yep, college professors get paid a fuckton to badly explain concepts to a small number of students... which they have to repeat for every new batch of students. The education system is a wacky thing.

[@SufferingAddict88]

@@JD-jl4yy Well to be fair to the professors, their main job is research. Teaching is just a side-job they have to do.

[@JD-jl4yy]

@@SufferingAddict88 exactly, and that doesn't always mix well. Having to spend a lot of time begging for grant money, organizing classes and (ineffectively) teaching. Doesn't seem like a very focused use of your time as a researcher...

[@kenken3000]

what is the prerequisite for understanding the likelihood ratio? I don't really understand the above explanation and how the LR is derived as it is.

[@crownedove]

most simple and clear explanation i see, thx for the great help! very easy to understand.

[@SpartacanUsuals]

Hi, glad to hear that it was helpful! Best, Ben

[@kelbyclements4470]

Hey Ben, can you comment on why you need to do this test? What is it most useful for? Model selection?

[@yasinmtk7284]

Ben, you are a legend.

[@Maha_s1999]

This is an invaluable channel!!

[@soodehmoallemian]

It was so help full! Thank you very much! looking for your new videos...

[@SpartacanUsuals]

Hi, thanks for your message and kind words. Glad to hear it was helpful! Best, Ben

[@soodehmoallemian]

Ben Lambert : It really was. And Thank you again :)

[@monishadamodaran677]

Could anyone explain about the chi square distribution thing from 5:40 mins???

[@ramansharma4025]

What it says is that for a very small degree of freedom such as 1 in this case the Chi-square distribution values are bunched towards value '0'.If you run LR test on your model and,however, get a high Chi-Sq value it is very unlikely to be obtained from a true Chi-Sq distribution. In other words,if Null Hypothesis ( i.e. the parameter Theta has no influence on response variable Y) is true you would not get a high Chi-Sq value from your test ( indicating that the predictor variable Theta actually had an impact on Y)

[@lizzyzhou4954]

@@ramansharma4025 I am also very confused by this chi square thing. If you check the distribution, with 1 degree of freedom, the pdf goes to infinity instead of zero.

[@rossbickmore8096]

You missed out a bracket.

[@drtshow777]

Excellent short lesson, thanks

[@SA-gt8lx]

Asking if null hypothesis is true. Is it the same as asking if theta0 is the true parameter? I dont quite get the null hypothesis thing.

[@freekevin91]

theta0 is always a true parameter. we ask whether the true parameter is equal to some particular value, say 0.

[@Mietchannel]

why is statistisc so confusing

[@anjalipatil2185]

🤣🤣🤣

[@qwertz6351]

Thanks A LOT! Greetings from Germany

[@datapprentice]

Can I use it to test for autocorrelation in a dynamic model?

[@SpartacanUsuals]

Hi, dependent on the circumstances yes. Best, Ben

[@HyperRefined]

The LR statistic is not necessarily asymptotically Chi squared distributed! In case of a random sample (i.e. iid random variables from some distribution F) this is always the case, true. But even in some i.n.i.d. cases (independent but not identically distributed) you need much stronger assumptions, usually regarding the Fisher Information, to secure the asymptotic Chi squared distribution

[@kylepena8908]

This is what i think needs to be fixed with Statistics. How many of us knew that? Why are there so many of these exceptions? There have got to be more general methods.

[@mashiathkhurshid8003]

Can this test be used for ARDL regressions?

[@filiphotti]

idk, what's up with your parenthesis? Should the whole expression be multiplied with 2 or just (log(L)|theta hat)

[@pirasa5]

Thank you! Excellent explanation!

[@michael052075]

Thanks for sharing . What software are you using to create your video lessons?

[@raichu4797]

Fantastic video! Thank you so much for making the contents so accessible! I wonder if there's a chart or something I should look at to make sure the value is high enough to reject the null hypothesis? is there a LR value chart regarding the level of freedom kind of thing?

[@tutululu8919]

thanks its so helpful

[@ranwei8070]

Great video! Thank u!

[@SpartacanUsuals]

Hi, thanks for your message, and kind words. Glad to hear that it was useful. Best, Ben

[@octale9445]

Nice explanation thank you

[@SpartacanUsuals]

Hi, glad to hear it helped. Best, Ben

[@cutieb32]

Shouldn't the 2 be negative?

[@lastua8562]

I don't think so. Under the null, LR has ChiSq distribution, which is always positive (as I have learned it).

[@regularviewer1682]

These videos are amazing but why so quiet!!

[@henrynnajiuba9591]

wouldn't really call this an 'introduction'. Quite technical for people like myself looking for a Dummies version!

[@michaelleming9123]

nicely done

[@melissahahn8552]

Awesome! Thanks a lot!

[@starmotry3766]

Missing a 2 or bracket: LR = 2lnL1 - 2lnL0

[@animals0feel1pain2]

Thanks so much!

[@Ds7nike]

ASHUME

[@Junkie-NE14ABJ]

No fucking clue