Explaining logistic regression

Рет қаралды 16,389

Very Normal

Күн бұрын

Пікірлер: 37

@zyansheep Ай бұрын

1:35, woh... fancy animation! I'm really feeling the corporate aesthetic :)

@very-normal Ай бұрын

it was definitely HR approved

@MKhan-zo8xo Ай бұрын

brilliant, brilliant video, the way you explain the relationships between important concepts is gold to me

@RyeCA Ай бұрын

I have an exercise just on that topic, excellent timing!

@omarpacheco2756 Ай бұрын

Great video, keep it up! Every time you mention a next video I get anxious to see it

@dogol284 Ай бұрын

I don't know anything about statistics, but with production quality like this I was amazed by how few views and subscribers you have.

@Bola382 Ай бұрын

the normal distribution of maximum likelihood estimators is an asymptotic result

@very-normal Ай бұрын

yee

@pfizerpflanze Ай бұрын

@@very-normalAnd only in a regular context, namely: - θ ∈ ℝ^d, d finite and constant - l(θ; y) ∈ C³ - sample space not depending on the parameter they are often neglected, but they are important also for the sake of glm, eg when using residual deviance for testing the goodness of fit. 🤓☝️ (waiting for your next video btw, glm is by far the hardest course this semester)

@KeyserTheRedBeard Ай бұрын

Great video, Very Normal. Looking forward to your next upload! I pressed that thumbs up icon on your content. Keep up the fantastic work. Your breakdown of logistic regression was incredibly insightful. How do you suggest addressing potential multicollinearity issues when interpreting the odds ratios in practical applications?

@nikolaylevkovsky9745 Ай бұрын

As always, good job communicating how statistical methods work! Your videos are easier to understand compared to most statistics content out there. Since you've been covering regression lately, do you plan on making one about Gaussian processes? They're one of my favorite tools in this field, but I'm not sure if you use that in biostatistics

@very-normal Ай бұрын

I’ve seen them used in some interesting applications for clinical trials. It could definitely be an interesting topic for a future video!

@nikolaylevkovsky9745 Ай бұрын

@very-normal They're a nice way of building a model with the same set of biases as you have, so I definitely didn't expect them to be used in trials 😅 Anyways, I'll be glad if you ever cover them. Touch on MLE and we're all set!

@tuongnguyen9391 Ай бұрын

You should make a video on how to explain concept. You have done a great job and I want to learn from you.

@olivierchabot Ай бұрын

Good stuff. Keep it up!

@BharmaArc Ай бұрын

Loved the video in general, but the use of pi instead of p honestly did confuse and distract me

@Louis-dj1dk Ай бұрын

Great stuff !

@pavanpreetgandhi6763 Ай бұрын

Looking forward to the next one one GLMs!

@komethtauch5151 Ай бұрын

you're actually the goat

@g-whittington Ай бұрын

Great as always

@mfabulencia Ай бұрын

what a time to be alive

@antonvanetten9071 Ай бұрын

Excellent

@galenseilis5971 Ай бұрын

Under typical assumptions of IID samples I can see how we get to asymptotically normal sampling distributions for estimates from MLE, and for sufficiently large samples we will have estimates with sampling distributions that are practically close enough to normal. It is not true in general that estimates from MLE will follow a normal distribution.

@very-normal Ай бұрын

Yea you’re right. I’m speaking from the perspective that we can use those typical assumptions, not from a more general one