Ahhh, there are like a million Data/ML channels but this is still the only one I subscribe to (after being burned a bit by Siraj). Love this guys ability to articulate complex phenomena in a way that makes sense.
Counts are often distributed under a Poisson. The domain is very important to mention when selecting a distribution, and the discrepancy between probability mass functions and density functions. Keep it up man, love your stuff.
Is there a way to create a custom probability distribution from a sample dataset? It can then generate new data with similar characteristics while remaining completely continuous?
Yes that is possible. In python, scipy has distributions where you call a "fit" function and pass in sample data. For example, check out scipy.beta.fit.
@@CodeEmporium Man, I feel like I'm really annoying here. I'm sorry! Be careful with this. Understand your data first: if there's any domain expertise you can throw into this, the data may be enforced to be a certain distribution, despite it not looking like it yet. For example: counting the number of times you see cars drive past your house within one hour blocks. Maybe you collect a handful of data. You notice a small tail at 2-5 cars, a peak at 7 cars, and a tail at 10 cars. You might think this is normal, but from the definition of the experiment, this is indeed a Poisson distribution: counting within set intervals.
I think only the normal distribution is technical here. The other 4 are a lot easier to pick up. Looking back, maybe could have easier explained the normal distribution. But I'll keep this mind for other videos
The real life examples for each of the distribution were amazing !!
Glad you liked them. Many thanks :)
1:00 Normal Distribution
5:37 Log-normal Distribution
7:30 Uniform Distribution
8:48 Beta Distribution
10:33 Chi-squared Distribution.
Thanks for the timestamps :)
Ahhh, there are like a million Data/ML channels but this is still the only one I subscribe to (after being burned a bit by Siraj). Love this guys ability to articulate complex phenomena in a way that makes sense.
Thanks for being a part of the community 🙂
@@CodeEmporium pleasure!
Counts are often distributed under a Poisson. The domain is very important to mention when selecting a distribution, and the discrepancy between probability mass functions and density functions. Keep it up man, love your stuff.
Thank you! More math videos to come!
Nice summary of five different topics that could be their own lessons
Thank you. Will def dive into these topics in thier videos in some consumable form. I just need to think of the best way to deliver this content
Hey buddy. Awesome as always. THANK YOU 💓
That was really helpful! Amazing content!
Many thanks and very glad you enjoyed it :)
You missed tweedie distribution which is used in insurance modelling
Man you are awesome!
how are y’all so smart… i left everything i learned about statistics back at where it started, at Uni :(
Honestly I did the same. But the more you work with this stuff on applications, the better you’ll remember it. :)
Is there a way to create a custom probability distribution from a sample dataset? It can then generate new data with similar characteristics while remaining completely continuous?
Yes that is possible. In python, scipy has distributions where you call a "fit" function and pass in sample data. For example, check out scipy.beta.fit.
@@CodeEmporium Thank you so much for the reply!
@@CodeEmporium Man, I feel like I'm really annoying here. I'm sorry! Be careful with this. Understand your data first: if there's any domain expertise you can throw into this, the data may be enforced to be a certain distribution, despite it not looking like it yet.
For example: counting the number of times you see cars drive past your house within one hour blocks. Maybe you collect a handful of data. You notice a small tail at 2-5 cars, a peak at 7 cars, and a tail at 10 cars. You might think this is normal, but from the definition of the experiment, this is indeed a Poisson distribution: counting within set intervals.
The sample means can still be normal even if the samples arent
Weibull gang stand up!
You have been heard
Data is is just a game of giving 100 different fancy names for the same concept to make it Extremely confusing for learners
sahi mai bhai
start at 1:00
i think 4 & 5 needed much more details, as much as we got for 1. but good video, thank you
poison ooops we need to talk about the poisson distribution as well
Another video for sure
Really good content
Thank you :)
good vid :)
My man, I love your videos, but the audio is often out of sync, just a heads up
Yep. Thanks for the heads up. I'm trying to get better with this for future videos :)
does anyone else see a lag between audio and video?
Sorry about that. It happens a couple of times through the video. Will try to correct for future videos
@@CodeEmporium no worries! :) I just couldn't tell if the issue was my computer or the video itself hahaha
no gamma :(
R u lipsing bro
Nah. It's your imagination
@@CodeEmporium thanks for replying.... Was watching sm of ur videos ...awseome stuff...thanks!!
One those videos where it's implicitly assumes that you know stats before hand and explicitly follow that assumption throughout the video...
I think only the normal distribution is technical here. The other 4 are a lot easier to pick up. Looking back, maybe could have easier explained the normal distribution. But I'll keep this mind for other videos
when i hear u first time its very weird u r voice does not match u. means don't know why its feels like that u r lisping and someone else is talking