Over the years, I have searched literally dozens of text books and articles to get an idea why the exponential distribution is a declining curve. This is the first instance that I have encountered a 'success' -- to use a statistical jargon. A similar reasoning explains the exponential smoothing model for forecasting, and only a couple of authors have really bothered to explain it. Great job Justin! Pretty soon, I guess you will need to revise the number of visits to your website!!!!! Thanks a lot!
This channel gets me some internel confidence that the topic I am searching for hours on the internet *will* be resolved with more than enough depth with the clarity needed.
Fantastic, zed statistics! This should be the number 1 option for explaining this topic out there! This is awesome (and what learning should be like). Thanks!
Justin explains exactly what I was wondering about the concept, or the big picture, about Exponential Distribution. I wanted so badly to interpret its graph, but there was no tutorial that told me about it until I reached this video. And this one is amazing! It just enlightens all that I wanted to know about this subject. Thanks a lot, Justin!
Another way to get an intuition for the shape of the exponential distribution would be to draw events on a number line you first draw them equal width apart (if it’s 3 hours per event then draw them one hour apart). Now sample 1 point per hour or something like that, you’ll see that the waiting times follow a uniform distribution. Now we can try to “randomize” the intervals a bit aka move the events around by for example one event 2 hours early and another 2 hours late to balance it out (so that the average rate stays the same). You can see that for the two intervals surrounding the event that’s moved two hours early, they were originally both 3 hours. Then, after the move, they become 1 and 5 hours. For the first interval, all waiting times within 1 hour still remain, on the other hand, higher waiting times between 1 and 3 hours are stripped away and converted to waiting times 3-5 hours in the second intervals. Higher waiting times have a higher chance of being converted to even higher waiting times, but lower waiting times do not. That’s why the density is higher towards shorter waiting times. I hope it makes sense. Another even simpler way to look at it is: if we sample the waiting times once per hour, for every waiting time of 3 hours, there MUST be one sample each for 2, 1 and 0 hours between it and the next event. On the other hand, if you have a waiting time of 1 hour, there isn’t a guarantee that there exist waiting times higher than 1 hour. In general terms, an instance of a longer waiting time corresponds to one instance each of all the waiting times shorter than it; however, the opposite doesn’t hold true (an instance of a shorter waiting time doesn’t guarantee an instance of any higher waiting time). That’s why the density HAS TO decrease towards higher waiting times.
This video and this channel are definitely the statistics explained in an intuitive way at its best. Love it and feel fortunate to find this resource. THANK YOU!
The axes on the graphs could do with some explanation... 6:06 On the Poisson distribution PMF graph on the left: - The X axis represents unique visitors to the website per hour. - The Y axis represents the probability of each discrete number of people visiting per hour. On the Exponential distribution PDF graph on the right: - The X axis represents hours until next arrival. - The Y axis does NOT represent the probability itself, which would have a scale of 0 to 1. Rather, the Y axis represents the PROBABILITY DENSITY, which is the RELATIVE LIKELIHOOD of each value on the X axis occurring. It's scale (0 to 3 in this case) is such that the total area under the graph = 1. 08:27 - On both CDF graphs, the Y axes DO represent the probability (scale 0 - 1). 10:21 until the end - The Y axis still represents the probability density (converted for minutes) and not the actual probability. 17:10 The explanation is a bit misleading. It doesn't explain why the graph falls; if the Y axis represented the probability of visitors arriving within discrete periods on the X axis, it would fall anyway, in a linear fashion, so that the product of the values on the X and Y axes remained uniform. But it does explain why the graph is CONCAVE, due the exponential nature of the function, and not linear. It's also unfortunate and confusing in this example that the PROBABILITY DENSITY at 0 minutes (0.05) is the same figure as the PROBABILITY that a visitor lands within each minute (0.05). They are not the same thing.
Because 0.95 keeps getting multiplied by itself in the function. In other words, it is a constant being raised to a power, which is the nature of an exponential function.
The last problem was just a fantastic one. First you treat it as an exponential distribution, so the probability of within one min becomes your probability of success. Then you treat it as geometric distribution. Brilliant!
You seriously rock! I have a test in a few days, and I have watched all of your videos regarding probability distributions. Feeling much much better! Again, thanks so much :)
Sir you are sooo kind person, you didn't let us to watch the entire poisson distribution video unlike many youtubers who take advantage of this and make viewers watch multiple videos, Sir you are super. Namaskaram sir🙏🙏🙏🙏🙏
Saving lives. My lecturer and textbook use lambda as both the Poisson mean and Exponential mean. Can't begin to explain how many hours I wasted not realising they were referring to two different means. Thought I was losing it. Was ready to drop out of math and try my luck in humanities.
This is a kind request to have a video series on Permutation, Combination ,Probability and Calculas. I must say your videos are very awesome. The way you explained things is fantastic. Thanks Justin
RUclips algorithms must be pretty good that it didn’t take me long to find this video on exponential distribution >< This one answered my question exactly which is why the exponential pdf looks like the way it does. Took me to click on 4 different videos and maybe 20mins of watching in total to get to this one
Thanks for the wonderful explanation. Just one confusion in the section "Visualisation (PDF and CDF)" - the Exponential distribution graph at @6:35 minutes is correct? because on the Y-axis you have put values greater than 1. but shouldn't these values be less than 1 representing the probability?
Would have been nice to state that the y-axis on the exponential dist is lambda for the PDF and a percentage for the CDF. Unlike the Poisson Dist as both are in percentage. This confused me as I wasn't sure what the Y axis meant. I naturally thought percentage and was wondering why nothing was adding up correctly especially at 16:44 - I was like, it should equal 0.025 or 2.5% which is of course wrong. I watched the whole video with the wrong assumption haha
- The Y axis on the exponential distribution PDF does not represent Lambda (nor the probability). It actually represents the PROBABILITY DENSITY, which is the RELATIVE LIKELIHOOD of each value on the X axis occurring. It's scale (0 to 3 in this case) is such that the total area under the graph = 1. But you're right that this was not explained at all in the video. - The Y axis on the exponential CDF and the Poisson distributions is probability, on a scale of 0 to 1, and not percentage, which would have a scale of 0 to 100.
This video really helped me a lot understanding the difference between Poisson and Exponential distributions. Outstanding ❤ Thank you and keep up the good work 🙏🏻
Good day! I see many other channels explaining this all wrong. You explain the poison mean as the inverse of the exponential mean and vice versa. The inverse of the exponential mean is also lambda in the exponential distribution function. Other channels are failing to explain that correlation.
Prob that visitor lands before 6:01 and before 6:21 are the same due to memorylessness. When applying the same logic to the problem you solved last, I don't get the logic behind the probabilities differing. Those should also be the same using the same logic and memorylessness?
thank you sir, great explanations and really helpful 😅😅👌👍 though in between the moments i do notice certain use of of rough language, just an advise on what could make these better. Personally i really like the way Mr. Grant on 3B1B talks, utterly admiring the beauty of the subject.😅😊 (to be quite precise, the beauties of geometrical patterns in curves of graphs and sequences and series that make them look the way they do, shall never be compared to a can of worms in my opinion, i am sorry)
Really fantastic! I know this distribution better than ever! btw, can you teach two more distribution - the gamma and the beta distribution. Thank you so much for your explanation anyway😄!
Hello, First I would like to express my appreciation and admiration for the epic way you're teaching these topics with a big time THANK YOU. I do want to ask this question pertaining to the Poisson requirement that the events must occur at a constant rate paradox. If they're occuring at a constant rate. Does this requirement apply on the average sense? Otherwise, if the rate of events (events per time) is constant, then why are what is the purpose of the distribution?
The Y axis represents the PROBABILITY DENSITY, which is the RELATIVE LIKELIHOOD of each value on the X axis occurring. It's scale (0 to 3 in this case) is such that the total area under the graph = 1.
Wait so based on the ending, does that mean that the longer his website goes without a visitor the less and less likely that his website will be visited, such that by 120 minutes or 2 hours it is almost 0% likely that someone will visit his website??
The example ends at 1 hour only.... so he is taking that in reference that after 1 or 2 hours there will be no visitor. Obviously, as time passes chances getting lower that visitor will come to your site.
munchees No, the conclusion which you seem to be drawing is incorrect. First, remember that the event i.e. arriving of visitors is independent. Second, note that the random variable X under consideration is the waiting time and not counting the number/probability of visitors arriving. So, to your question, yes the probability of visitor arriving in the 120th minute (and after) is almost 0, but it also means that it is almost 100% likely that the visitor will arrive much before than 120 mins.
In the last part where he mentioned that first visitor lands within 2nd minute y dont we just take x as 2 and apply the formula? Y do we consider it as two distinct events and multiply it with 0.95 ?
You could take x as 2 and apply the formula, but he's doing it the long way to try to explain the shape of the graph. He's actually explaining why the graph is exponentially concave (not why it falls).
This channel is underrated.
Over the years, I have searched literally dozens of text books and articles to get an idea why the exponential distribution is a declining curve. This is the first instance that I have encountered a 'success' -- to use a statistical jargon. A similar reasoning explains the exponential smoothing model for forecasting, and only a couple of authors have really bothered to explain it. Great job Justin! Pretty soon, I guess you will need to revise the number of visits to your website!!!!! Thanks a lot!
All of your videos keep giving me the Eureka moment at some point in the video. Keep doing what you're doing ZED. Lots of love and admiration.
This channel gets me some internel confidence that the topic I am searching for hours on the internet *will* be resolved with more than enough depth with the clarity needed.
Fantastic, zed statistics! This should be the number 1 option for explaining this topic out there! This is awesome (and what learning should be like). Thanks!
Thanks edu boss! Share it round! :)
The best intuitive video on exponential distribution I have seen so far.. Thanks Justin for sharing.
Best I've seen by far
your voice so soothing bruh, it plug all the theories into my head perfectly
Brilliant teacher , very clear with a commonsense approach.
Justin explains exactly what I was wondering about the concept, or the big picture, about Exponential Distribution. I wanted so badly to interpret its graph, but there was no tutorial that told me about it until I reached this video. And this one is amazing! It just enlightens all that I wanted to know about this subject. Thanks a lot, Justin!
Thank you just soooo much! May the lord give you paradise in this in this one and afterlife.
I swear bro you are one of the best teachers out there!
All I can say is Thank you from the bottom of my heart.... This saved me...
Another way to get an intuition for the shape of the exponential distribution would be to draw events on a number line you first draw them equal width apart (if it’s 3 hours per event then draw them one hour apart). Now sample 1 point per hour or something like that, you’ll see that the waiting times follow a uniform distribution. Now we can try to “randomize” the intervals a bit aka move the events around by for example one event 2 hours early and another 2 hours late to balance it out (so that the average rate stays the same). You can see that for the two intervals surrounding the event that’s moved two hours early, they were originally both 3 hours. Then, after the move, they become 1 and 5 hours. For the first interval, all waiting times within 1 hour still remain, on the other hand, higher waiting times between 1 and 3 hours are stripped away and converted to waiting times 3-5 hours in the second intervals. Higher waiting times have a higher chance of being converted to even higher waiting times, but lower waiting times do not. That’s why the density is higher towards shorter waiting times. I hope it makes sense.
Another even simpler way to look at it is: if we sample the waiting times once per hour, for every waiting time of 3 hours, there MUST be one sample each for 2, 1 and 0 hours between it and the next event. On the other hand, if you have a waiting time of 1 hour, there isn’t a guarantee that there exist waiting times higher than 1 hour. In general terms, an instance of a longer waiting time corresponds to one instance each of all the waiting times shorter than it; however, the opposite doesn’t hold true (an instance of a shorter waiting time doesn’t guarantee an instance of any higher waiting time). That’s why the density HAS TO decrease towards higher waiting times.
This video and this channel are definitely the statistics explained in an intuitive way at its best. Love it and feel fortunate to find this resource. THANK YOU!
The axes on the graphs could do with some explanation...
6:06 On the Poisson distribution PMF graph on the left:
- The X axis represents unique visitors to the website per hour.
- The Y axis represents the probability of each discrete number of people visiting per hour.
On the Exponential distribution PDF graph on the right:
- The X axis represents hours until next arrival.
- The Y axis does NOT represent the probability itself, which would have a scale of 0 to 1. Rather, the Y axis represents the PROBABILITY DENSITY, which is the RELATIVE LIKELIHOOD of each value on the X axis occurring. It's scale (0 to 3 in this case) is such that the total area under the graph = 1.
08:27 - On both CDF graphs, the Y axes DO represent the probability (scale 0 - 1).
10:21 until the end - The Y axis still represents the probability density (converted for minutes) and not the actual probability.
17:10 The explanation is a bit misleading. It doesn't explain why the graph falls; if the Y axis represented the probability of visitors arriving within discrete periods on the X axis, it would fall anyway, in a linear fashion, so that the product of the values on the X and Y axes remained uniform. But it does explain why the graph is CONCAVE, due the exponential nature of the function, and not linear. It's also unfortunate and confusing in this example that the PROBABILITY DENSITY at 0 minutes (0.05) is the same figure as the PROBABILITY that a visitor lands within each minute (0.05). They are not the same thing.
Thank you Nick! ❤
@@rom3o.s_regr3t You are welcome, Bontle :)
pois(X) and exp(X) bless you sir for this great lecture. Wonderful.
the last few minutes gave the most important intuition! Thanks! 17:05 Why is it called "Exponential"??
Because 0.95 keeps getting multiplied by itself in the function. In other words, it is a constant being raised to a power, which is the nature of an exponential function.
Best content for learning statistics for data science
Clearest stats video I have ever watched. Thank you
The last problem was just a fantastic one. First you treat it as an exponential distribution, so the probability of within one min becomes your probability of success. Then you treat it as geometric distribution. Brilliant!
Best channel for Statistics!!!
You seriously rock! I have a test in a few days, and I have watched all of your videos regarding probability distributions. Feeling much much better! Again, thanks so much :)
Omg, cant believe this video doesnt have more likes! top level sta video!
Sir you are sooo kind person, you didn't let us to watch the entire poisson distribution video unlike many youtubers who take advantage of this and make viewers watch multiple videos, Sir you are super. Namaskaram sir🙏🙏🙏🙏🙏
Saving lives. My lecturer and textbook use lambda as both the Poisson mean and Exponential mean. Can't begin to explain how many hours I wasted not realising they were referring to two different means. Thought I was losing it. Was ready to drop out of math and try my luck in humanities.
I understand how simple it is just because of your this video. Thank you so much.
Brilliant, loved the simple PDF explanation at the end
The way you explained why the pdf looks like it is really amazing! Thank you! I finally realized exponential is related to binomial distribution!
Man i would have never understood it any other way. Outstanding explanation 👏👏👏
These videos are incredibly informative ! I encourage you make some more !!!
Amazing Class! Salute from Brazil.
This is a kind request to have a video series on Permutation, Combination ,Probability and Calculas. I must say your videos are very awesome. The way you explained things is fantastic. Thanks Justin
You are so good in explaining maths.
Best intuitive explanation I’ve found. Thanks!
you re video is just perfect. you also explain very well why things are like this or like that
RUclips algorithms must be pretty good that it didn’t take me long to find this video on exponential distribution >< This one answered my question exactly which is why the exponential pdf looks like the way it does. Took me to click on 4 different videos and maybe 20mins of watching in total to get to this one
Thanks for the wonderful explanation. Just one confusion in the section "Visualisation (PDF and CDF)" - the Exponential distribution graph at @6:35 minutes is correct? because on the Y-axis you have put values greater than 1. but shouldn't these values be less than 1 representing the probability?
Wow. Just wow. This video is marvellous! We really appreciate your effort!
This video is amazing the only video which explains exponential distribution in depth . Thankyou so much
This really helps me understand how the statistical tests built on these distribution works!
The best video for understanding exp dist...loved the way it explains!
Best explanations ever. Thanks.
Would have been nice to state that the y-axis on the exponential dist is lambda for the PDF and a percentage for the CDF.
Unlike the Poisson Dist as both are in percentage.
This confused me as I wasn't sure what the Y axis meant. I naturally thought percentage and was wondering why nothing was adding up correctly especially at 16:44 - I was like, it should equal 0.025 or 2.5% which is of course wrong. I watched the whole video with the wrong assumption haha
It's mentioned on the y axis, the values. So it's kinda self explanatory 😅
you are right my friend. I had the same doubt throughout the video
- The Y axis on the exponential distribution PDF does not represent Lambda (nor the probability). It actually represents the PROBABILITY DENSITY, which is the RELATIVE LIKELIHOOD of each value on the X axis occurring. It's scale (0 to 3 in this case) is such that the total area under the graph = 1. But you're right that this was not explained at all in the video.
- The Y axis on the exponential CDF and the Poisson distributions is probability, on a scale of 0 to 1, and not percentage, which would have a scale of 0 to 100.
Great explanation. Cannot be better than that. Crystal clear my concept. Thanks
Awesome. Especially, the last sections explanation was crystal clear. Thank you.
Appreciating your smart way to lead us through exponential distribution
After seeking around a lot of videos its the only video which shows why its PDF looks the way it looks
wow! really good explaination
Interesting. when you were explaining the pdf, I couldn't help but notice that behavior was similar to the geometric distribution. I wonder why.
I think one of the best explanations on Exponential Distribution. Could you please share any content with its link to CTMC and Transient Analysis.
You are an incredible instructor.
Awesome explanation, Sir
He's the teacher we never had.
Ohh man you made me very clear on exponential distribution thank you so much for it . Also please make a video on Gamma distribution
This video really helped me a lot understanding the difference between Poisson and Exponential distributions. Outstanding ❤ Thank you and keep up the good work 🙏🏻
Thank u so much!These lectures are very intutive!!
amazing videos. your explanations, oration, recording, and visuals all are superb!
archangel of stats explanations thx zed
Thank You so much for the explanation in the "exactly" scenario, zedstatistics. This helped me a lot. Thanks a million.
very very clear explanation. Thank so much. You did help me to understand Possion and Exponential!
This topic was explained very nicely. Thank you.
Good day! I see many other channels explaining this all wrong. You explain the poison mean as the inverse of the exponential mean and vice versa. The inverse of the exponential mean is also lambda in the exponential distribution function. Other channels are failing to explain that correlation.
Prob that visitor lands before 6:01 and before 6:21 are the same due to memorylessness. When applying the same logic to the problem you solved last, I don't get the logic behind the probabilities differing. Those should also be the same using the same logic and memorylessness?
thank you sir, great explanations and really helpful 😅😅👌👍
though in between the moments i do notice certain use of of rough language, just an advise on what could make these better. Personally i really like the way Mr. Grant on 3B1B talks, utterly admiring the beauty of the subject.😅😊
(to be quite precise, the beauties of geometrical patterns in curves of graphs and sequences and series that make them look the way they do, shall never be compared to a can of worms in my opinion, i am sorry)
Great video thanks for the help!
Simply superb, thanks for making these videos. Hope you keep making more videos on statistics!
Hello Sir, I have watched many of your vedios..And I really like those.. Kindly make one vedio on endowgenity. or suggest me some source.
Thank you.🙂
Wow this is soo coool! It is a great addition to "Practical statistics for data scientists" book. Thanks!
The last part reminds me of the binomial distribution without de combinations in the formula.
Damn that's awesome! Now i understand where the ' exponential' came from.🎉
Did I just learn what exponential distribution is? :)
Thank you!
your explanations are really great. could you do more distrubution videos
I always thumps up before watching you're videos :p
so cool, I wondered why distribution looks like that. so clear now!
Really fantastic! I know this distribution better than ever! btw, can you teach two more distribution - the gamma and the beta distribution. Thank you so much for your explanation anyway😄!
Okay, okay, so anyone would listen to Justin explain even how sand is made. Thanks for the video !
Sand videos, hey? That's really gonna take my channel in a different direction but let's do it!
Hi, could you make a video about Gamma distribution? Thanks
I cannot thank you enough for this video
Hello, First I would like to express my appreciation and admiration for the epic way you're teaching these topics with a big time THANK YOU. I do want to ask this question pertaining to the Poisson requirement that the events must occur at a constant rate paradox. If they're occuring at a constant rate. Does this requirement apply on the average sense? Otherwise, if the rate of events (events per time) is constant, then why are what is the purpose of the distribution?
Wow. That is all I can say.
It’s all so clear now😌
This was really helpful! Thanks a lot for your kind effort.
@7:36 what is the unit on the Y-axis of the graph to the right?
The Y axis represents the PROBABILITY DENSITY, which is the RELATIVE LIKELIHOOD of each value on the X axis occurring. It's scale (0 to 3 in this case) is such that the total area under the graph = 1.
Fantastic video. Keep it up.
In getting the probability for the visitor to land within third minute, why dont we just do P(X
Thankyouuuu so much!❤❤, Very well explained
Fantastic explanation.
thank you so much for the explanation on exponential distribution i found it easy to understand
Thank you so much. This was very well explained.
Fantastic. Keep up your good work!
Wait so based on the ending, does that mean that the longer his website goes without a visitor the less and less likely that his website will be visited, such that by 120 minutes or 2 hours it is almost 0% likely that someone will visit his website??
The example ends at 1 hour only.... so he is taking that in reference that after 1 or 2 hours there will be no visitor. Obviously, as time passes chances getting lower that visitor will come to your site.
munchees No, the conclusion which you seem to be drawing is incorrect. First, remember that the event i.e. arriving of visitors is independent. Second, note that the random variable X under consideration is the waiting time and not counting the number/probability of visitors arriving.
So, to your question, yes the probability of visitor arriving in the 120th minute (and after) is almost 0, but it also means that it is almost 100% likely that the visitor will arrive much before than 120 mins.
Hi thanks for making these videos, can you make one such video on Kappa values and Weibull distribution
Thank you very helpful, can you please do a video on gamma distributions
Sir, thank you so much for the very clear lesson :)
In the last part where he mentioned that first visitor lands within 2nd minute y dont we just take x as 2 and apply the formula? Y do we consider it as two distinct events and multiply it with 0.95 ?
You could take x as 2 and apply the formula, but he's doing it the long way to try to explain the shape of the graph. He's actually explaining why the graph is exponentially concave (not why it falls).
Does it make sense to look at the probability of an event occurring between two points for an exponential distribution?
YOU'RE THE BEST OH MY GOD. THANK YOU
Amazing explanation
3:00 The horse kick deaths were decreasing by the same ratio as soldiers got run over by cars.
Thanks a lot. That really is an awesome explanation
Sir, can you please explain random variable to Probability distribution function of Continuous case.