Proof that the Binomial Distribution tends to the Poisson Distribution

Поделиться
HTML-код
  • Опубликовано: 3 ноя 2024

Комментарии • 142

  • @khcrafts1972
    @khcrafts1972 6 лет назад +61

    I don't know who you are, but I will look for you, I will find you, and I'll appreciate you in person! You're a lifesaver!

  • @NA-gx9cr
    @NA-gx9cr 3 года назад +21

    I am speechless at how well this is explained.

    • @jbstatistics
      @jbstatistics  3 года назад +7

      Thank you for the very kind words!

  • @baobui3002
    @baobui3002 5 месяцев назад +5

    You are the biggest nerd with the best teaching skill ever!

    • @jbstatistics
      @jbstatistics  5 месяцев назад

      I don't know what to say, other than it sounds true to me :)

  • @user-se9uk2py5k
    @user-se9uk2py5k 4 года назад +4

    Excellent. Following the complete series of videos. I think you are really generous sharing all of these good material. Thank you from Spain!

  • @TrippinCozy
    @TrippinCozy 6 лет назад +3

    Been searching for 2hours for this kind of thing !! Thaaank God we have access to information and there are people like you , otherwise, university would have been impossible!

  • @allstringsattached
    @allstringsattached 8 лет назад +4

    SUPERB Explanation!!!! I was getting so stuck using Stirling's formula. Your explanation is so crisp and effective. Thanks so much for the efforts.

    • @jbstatistics
      @jbstatistics  8 лет назад

      You are very welcome! Thanks for the compliment!

  • @jozokukavica9814
    @jozokukavica9814 6 лет назад +3

    Your tone and enthusiasm takes this proof to a whole another level. It's so good !!!

  • @janseyttacheung2952
    @janseyttacheung2952 10 лет назад +45

    1:56 in the last step it should be 1-(x-1)/n not 1-(x+1)/n. it won't affect the result though

    • @mikemazanetz4183
      @mikemazanetz4183 Год назад

      not seeing what you mean? (n/n)-(x+1)/n = 1 - (x+1)/n which is what he has.

    • @thenerdguy9985
      @thenerdguy9985 8 месяцев назад +1

      @@mikemazanetz4183 It's not (n-(x+1)). It is (n - x + 1). I hope you get it now.

    • @yinusao.emmanuel8378
      @yinusao.emmanuel8378 7 месяцев назад +2

      You're right and yes, it won't affect the final result

  • @jbstatistics
    @jbstatistics  11 лет назад +1

    I don't know what you mean. I have a video on approximating binomial probabilities using the Poisson distribution. If that's what you're looking for, it's in my discrete probability distributions playlist. If that's not what you're looking for, you'll have to try to express it in a different way. Cheers.

  • @philandthai
    @philandthai 5 лет назад +4

    Thank you Dr. Balka. Your videos are models of clarity (and I like the Canadian references too!). Like Mozart’s piano concertos, there is not a note out of place. I really appreciate the effort you have put into producing them and I hope you get academic credit for your work educating the general public. I am filled with admiration for the person who originally derived this proof.

    • @jbstatistics
      @jbstatistics  5 лет назад +3

      Thanks for the kind words Phil!

    • @mikemazanetz4183
      @mikemazanetz4183 Год назад

      that would be Simeon-Denis Poisson in 1837 who introduced it in a scientific text on the subject of jury verdicts in criminal trials, no doubt

  • @yaweli2968
    @yaweli2968 Год назад +1

    I love how you teach with enthusiasm. Amazing!Thanks.

  • @FBWalshyFTW
    @FBWalshyFTW 11 лет назад +9

    Wonderful, wonderful proof.

  • @anuskasaha8117
    @anuskasaha8117 2 года назад +3

    I usually don't comment but you really made this hard one so simple thank you so much it really cleared all of my doubt

  • @wayneli5586
    @wayneli5586 8 лет назад +1

    Extremely helpful! Prof didn't talk about the proof in the class, thank you so much!

  • @mazvta
    @mazvta 5 лет назад +11

    The animation is his voice from 5:06 onwards is my drug

    • @jbstatistics
      @jbstatistics  5 лет назад +2

      It's a very exciting part of the proof!

  • @MogauTau
    @MogauTau 4 месяца назад

    This video changed my life...thanks brother

  • @flyingbirds6794
    @flyingbirds6794 Год назад

    Absolutely world-class explanation,I could understand even without the sound of the video.

  • @dmitriidoroftei4478
    @dmitriidoroftei4478 2 года назад +2

    Wow, nice proof. I really liked the way you've explained it :D

  • @Botazz
    @Botazz 12 лет назад

    I have my final tomorrow. You don't know how thankful I am :) Thank you very much sir!

  • @yooos3
    @yooos3 11 лет назад

    The Best Explanation.
    I love your videos because the book I have has crap.
    Thank you so much !

  • @hash6129
    @hash6129 6 месяцев назад

    I am just going to say WOW!!!
    This was an easy explanation!!!

  • @nemdenemam9753
    @nemdenemam9753 4 года назад +1

    ok, this series is amazing. I wish I found your channel earlier.

  • @moomoomilk95
    @moomoomilk95 8 лет назад +56

    God, is that you?

    • @jbstatistics
      @jbstatistics  8 лет назад +18

      I'm just one of the messengers, trying to share what little I know about this universe :)

    • @alaaabdo5163
      @alaaabdo5163 6 лет назад +1

      all respect

  • @thomgrunauer6526
    @thomgrunauer6526 4 года назад

    I like how you get super excited at the end of the proof

  • @teetawatchaikree7421
    @teetawatchaikree7421 3 года назад

    you're vid are as clear as crystal Thanks!,

  • @NeelShah0211
    @NeelShah0211 3 года назад +5

    Amazing explanation. However I have a doubt.
    At 4:35 you said that :
    lim as n -> infinity for (1-lambda/n)^n is e^-lambda. However as n is going to infinity why doesn't it make (lambda/n) = 0 and leave us with 1^n just like the last term which became 1 at 4:52?

    • @mysigt_
      @mysigt_ Год назад

      Notice that in the second term, n is the exponent, and it tends to infinity. While in the last term, the exponent is -x, which is fixed. You might think of it as comparing how fast the inside approaches 1 to how fast the outside grows; it tends toward e^-lambda as n approaches infinity.
      I’m aware that this is a two year old comment, but it might be helpful to others.

  • @liusuyun3901
    @liusuyun3901 8 лет назад +4

    at 3:00, the last term should be 1- (x-1)/n. Anyway, it doesn't matter. Your elaboration is so vivid. Thanks.

    • @jbstatistics
      @jbstatistics  8 лет назад +8

      +LIU Suyun Thanks for the compliment! Yes, I've always been very annoyed about that little error. I do my best to go error free, but it's tough to run the table every time. Fortunately the limit is the same with or without the error in the sign. I'll put up a clean version at some point. Cheers.

    • @YsmiraWoW
      @YsmiraWoW 8 лет назад

      Is there a clean version yet, or could you explain the error to me?.. I'm using this derivation in a project and would like that to be error free as well, even though it doesn't technically matter once n->∞)

    • @jbstatistics
      @jbstatistics  8 лет назад +4

      The term (n-x+1)/n is correct, but there's a mistake in the next line. (n-x+1)/n can be written as n/n + (-x+1)/n, or 1 - (x -1)/n. So the x + 1 in the numerator of the last fraction on that slide should be x - 1.

  • @mukamafrancois2776
    @mukamafrancois2776 Месяц назад

    Thank you so much! You are very highly talented!

  • @enesilbuga2568
    @enesilbuga2568 8 лет назад

    This is one hell of a explanation. If only my teachers could explain it like this.. Mind it if you give a visit to Amsterdam?

  • @estefaniakiara-elizabeth8538
    @estefaniakiara-elizabeth8538 6 лет назад

    Wow! Your explanation is REALLY AWESOME! Thank you so much for this wonderful video, it really helped me a lot.

  • @jbstatistics
    @jbstatistics  12 лет назад

    You are welcome. Best of luck on your final!

  • @beinghimself
    @beinghimself Год назад

    Ok here is the explanation and i know it’s 7 years later. You could imagine that the binomial distribution is for either happening or not happening and for n trials. Now we are doing the same but with a time period of one second. The probability of the « happening » in one second is lumbda (i’m referring to it as k) and you might say: yeah i can also find the probability for 1 hour or even for 1 millisecond or even for 1 moment, it would just depend on how many time intervals there are in a second or the opposite, then we could say k= np as p is the probability of the « happening in each of these time intervals contained in the second we are studying, now imagine that we consider that for every moment, there would be a probability of the event happening but which would be very small. So understanding that we are dividing the probability in a second or an hour or a minute by a very big number of very small time intervals is very crucial. The poisson process is basically a binomial process for a gigantic amount of very small time intervals where the probability is very small. So the limit of the binomial process formula when n is +oo and p goes to 0 is the formula for the binomial. Reason why we do this: when we are trying for a second, we are actually trying for every moment of it, with a small probability for every moment. Hope it helps, sorry i’m first year college and it’s the first time i see the video and comments ❤

  • @christophercrawford641
    @christophercrawford641 2 года назад

    Ths proof implicitly assumes that x

  • @musicfan1695
    @musicfan1695 4 года назад

    you explained that very well, but the only problem I faced was is that your speech goes with a whistle at the end of many words you speak , which annoys a bit. If you could do something about it , like using a voice modulator or not doing it (the whistling thing you do) , it would be very helpful

  • @jbstatistics
    @jbstatistics  11 лет назад

    Thanks! I'm glad you found it useful!

  • @masonmarsai
    @masonmarsai Месяц назад

    my man just broke down a hour's worth lecture in 5min!!!!! wow

  • @boutiquemaths
    @boutiquemaths Год назад

    Such a cool result and beautiful explanation, ty! I’m wondering when it makes sense that p changes.. how can it be that p goes to 0 as n increases? Isn’t that a bit bizaare? Oh.. maybe the more often something is done, there is a feedback loop that dampens the event from happening again as frequently. For example, if we measure the number of accidents on building sites, it could be quite high at the beginning, but the more buildings we build, and the more accidents we have, the more we learn from it, so p (where p = probability of accident), goes down. Is that a way to think of it?

    • @jbstatistics
      @jbstatistics  Год назад +1

      It's not quite like that. p isn't changing in any given practical scenario.
      The casual gist is that the for any fixed value of lambda = np, the approximation will get better as n gets larger. (If lambda is fixed at np, then as n -> infinity p -> 0.) Or, even shorter and more casually: the approximation works well when n is large and p is small.
      So it's not p changing in a practical scenario, it's that mathematically, the binomial distribution gets closer to the Poisson as n goes to infinity and, as a consequence of lambda = np staying fixed, p goes to 0. In practice, that means this approximation will work best in situations where n is large and p is small.

    • @boutiquemaths
      @boutiquemaths Год назад

      @@jbstatistics got it, ty.

  • @rdalbanus
    @rdalbanus 9 лет назад +3

    Really good explanation. Keep up!

  • @mingshc
    @mingshc 7 лет назад

    Thank you!!! This clears the brain fog I got from the textbook!

  • @brioche_al_cioccolato
    @brioche_al_cioccolato 2 года назад +1

    Nice video, one question...what do you mean by "np stays constant"? That it is a finite number (2,6,9...)?

    • @jbstatistics
      @jbstatistics  2 года назад +1

      np is a positive constant that doesn't change as n -> infinity. It might be 2, 7, 3.23422552, 212.6, etc. n is going to infinity and p is going to 0 in such a fashion that the value of np stays the same.

    • @brioche_al_cioccolato
      @brioche_al_cioccolato 2 года назад

      @@jbstatistics Thanks for your response, i have been banging my head against the wall for several days trying to figure out this distribution and this video saved me. I have one last (probably very stupid) question, what happens if lambda tends to infinity? I don't understand if it is still considered constant.

    • @duanas6409
      @duanas6409 Год назад

      @@brioche_al_cioccolato I think lambda doesn't go to infinity. lambda = np, which is constant

  • @deltax7159
    @deltax7159 11 месяцев назад

    you are doing Gods work.

  • @tse4620
    @tse4620 6 месяцев назад

    Thanks for your brilliant video! At 1:55, I think the last term in the last row should be (1- (x-1)/n) instead of (1- (x+1)/n) right?

    • @jbstatistics
      @jbstatistics  6 месяцев назад +1

      Yes. Many years ago I put an annotation over it to fix it, but those stopped being a thing. I hate that there's a typo there, but fortunately it doesn't matter in the end since the term goes to 0 either way. Sorry about imposing the extra mental effort!

    • @tse4620
      @tse4620 6 месяцев назад

      @@jbstatistics no problem and thanks for your prompt reply! Always support your great videos!

  • @sjjayaswal7621
    @sjjayaswal7621 2 года назад

    Good information

  • @FoobsTon
    @FoobsTon 6 лет назад

    i bust my ass looking at this stuff in Actuarial text books 20 years ago. Oh, to have had a great tutor explain it well on RUclips. Good stuff Sir.

    • @jbstatistics
      @jbstatistics  6 лет назад +1

      Thanks for the kind words! I'm glad to be of help!

    • @jbstatistics
      @jbstatistics  6 лет назад

      I'm glad to be of help! Thanks for the kind words!

  • @PreetanjaliRay
    @PreetanjaliRay 9 лет назад +1

    Great explanation! Thank u so much! :)

  • @TokunboHiamang
    @TokunboHiamang 11 лет назад

    This is so good. I don't even know where to start!

  • @alexc8289
    @alexc8289 7 лет назад +1

    Great explanation!

  • @Noxxy2445
    @Noxxy2445 7 лет назад +1

    Excellent explanation!!

  • @thekolbaska
    @thekolbaska 11 лет назад

    This is fantastic! Love it!

  • @cleverclover7
    @cleverclover7 4 года назад

    Magnificent!

  • @Victual88
    @Victual88 3 года назад

    Hello JB, thanks for this great proof, do you mind if I ask a question @2:18,
    I don't understand how n! is expressed at the end of the series, where (n-x+1)(n-x)!
    I get the step down of factorials, I just don't understand the introduction of the variable x in the series.
    (Since x is already used in the bottom row and therefore both must have the same value.) I think this is the connection I am missing.
    Thanks for your help and time
    Kind regards
    Connor

    • @iSMohapatra
      @iSMohapatra 2 года назад

      That's because the terms after (n-x)! gets cancelled out with the bottom term. They are just repeating in the numerator. Hope it helps.
      Every time I get questions like this I substitute them with numbers and that solves it for me. Try n as 7 and x as 3. It would give a clear picture.

  • @kevinfejer9281
    @kevinfejer9281 4 года назад

    men i love u! have literraly salved my semester

  • @retrodude123
    @retrodude123 5 лет назад

    You are an absolute G

  • @shivasmart967
    @shivasmart967 3 года назад +1

    tq sir

  • @eirikdegard4498
    @eirikdegard4498 7 лет назад

    thanx, just what i was looking for.

  • @krishnarpitadebnath9706
    @krishnarpitadebnath9706 4 года назад

    Can you show the proof of- Negative binomial distribution tends to normal distribution as k tends to infinity and p tends to infinity.
    Please help🙏

  • @Jacob930321
    @Jacob930321 10 лет назад

    0:43 I know the definition of e, and understand Single variable calculus (thus, how limits works and their proofs). How can I understand that you get e^x by popping in an x there?

    • @jbstatistics
      @jbstatistics  10 лет назад

      Hi Jacob. There is some discussion of this in the Wikipedia article "Characterizations of the Exponential Function", available at en.wikipedia.org/wiki/Characterizations_of_the_exponential_function. Cheers.

  • @xualexander8451
    @xualexander8451 9 лет назад +1

    Thx buddy, this really helps!

  • @TheMrAineas1
    @TheMrAineas1 3 года назад +1

    Great video. But I think it's simpler if we use the moment generating functions by showing that the Poisson MGF is the limit of the binomial MGF. Granted, you would still have to derive the MGFs but at least you wouldn't have to deal with limits of factorial expressions.

  • @dennisestenson7820
    @dennisestenson7820 10 месяцев назад

    Fantastic!

  • @tylerthomas9405
    @tylerthomas9405 5 лет назад

    Really helpful, Thanks! Instant Subscribe!!!

  • @raihannyathifa5177
    @raihannyathifa5177 3 года назад

    Thank you!

  • @fish7828
    @fish7828 3 года назад

    Such a wonderful video, as well as a proof, whoever discovered it needs a neck brace to handle that massive brain!

  • @jasminehussain7393
    @jasminehussain7393 8 лет назад

    Thank you so much sir

  • @ichigokurosaki3743
    @ichigokurosaki3743 Год назад +1

    Thankyou sir

  • @egyptianplanner
    @egyptianplanner Год назад

    Exceptional

  • @Ronnicus
    @Ronnicus 9 лет назад

    So is the poisson distribution just a special case of the binomial or is it a stand alone function that can be used to approximate the binomial in this special case?

    • @christophercrawford641
      @christophercrawford641 2 года назад

      Yes, it is the special case of the binomial distribution for n really large (but not p), so that any number of occurances are possible (but not probable) within a certain time limit. The binomial distribution limits the number of occurances to a maximum of n.

  • @yuhanhung6701
    @yuhanhung6701 10 месяцев назад

    Thank you.

  • @ДаниилСарафанников-л2е
    @ДаниилСарафанников-л2е 10 месяцев назад +1

    That’s great, thanks!

  • @zagros18
    @zagros18 11 лет назад

    can you please make a video that how can we separate binomial distribution from passion distribution, by taking an example pleaseeeeee,,,, i really need it

  • @jeffreya.faulkner8367
    @jeffreya.faulkner8367 3 года назад

    A little English issue: You are referring to factors as terms. Factors are multiplied by one another while terms are added to one another.

  • @manuzrp1
    @manuzrp1 3 года назад

    There is a typo: x+1 should be x-1

  • @wenqijie945
    @wenqijie945 9 лет назад

    (n-x+1)/n can be converted to n/n-(x+1)/n?

  • @Pruthvikajaykumar
    @Pruthvikajaykumar 3 года назад

    wow so much in 5 minutes, thank you

  • @chizpafea
    @chizpafea 9 лет назад

    I suppose that to proof that the negative binomial distribution tends to poisson distribution we'll keep the same steps!... now i have a lamda=n(1-p)... am i right?

    • @debasishkar761
      @debasishkar761 7 лет назад

      the mean and variance is equal in poisson distbn>>>and lamda=n*p*(1-p) whr p=success

  • @abhishekkumar-mm4ye
    @abhishekkumar-mm4ye Год назад

    Loved that

  • @umairalvi7382
    @umairalvi7382 6 лет назад

    Absolutely perfect 😀

  • @jeskow19
    @jeskow19 5 лет назад

    I'm just going to believe you and not worry about trying to prove this myself.

  • @ahmetcihan8025
    @ahmetcihan8025 4 года назад

    Masterpiece Thanks sır.

  • @weihengzhang3618
    @weihengzhang3618 3 года назад

    You saved my life!!!😂

  • @shreejanshrestha1931
    @shreejanshrestha1931 3 года назад

    at 5:13, damn that stress he put in saying poisson haha !! puuuuuwaaaaasuuuuuuuuuu distrubution...

  • @luisfernando262
    @luisfernando262 2 года назад

    How can we say that p tends to 0? Shouldn't p be a constant describing the success probability of independent trials? Thus being fixed.
    Edit: p to 0, not infinity

    • @jbstatistics
      @jbstatistics  2 года назад +1

      I'm not sure what you mean. First, p isn't going to infinity, it's going to 0. Second, in some situations p is 0.8, in some 0.4, in some 0.2, in some 0.1, in some 0.05, in some 0.025, and so on. The value of p depends on the situation. In some situations it's small. Now let's suppose we're thinking about situations in which p is small, then another situation where p is smaller still, then another where p is smaller still, and so on.
      We're not saying suppose we're rolling a fair die a bunch of time and counting up the number of sixes, then pretending that probability is going to 0. in practical situations, related to this video topic, we are considering a situation where p is very small.

    • @luisfernando262
      @luisfernando262 2 года назад

      @@jbstatistics yes, i meant 0, my mistake. So basically this is just saying that the approximation is most accurate for rare bernoulli success and high amount of trials ?

    • @jbstatistics
      @jbstatistics  2 года назад

      @@luisfernando262 Yes, pretty much.

  • @bablumehta8435
    @bablumehta8435 7 лет назад

    sir please give a prove about as normal distribution as a limiting form of binomial distribution
    plz sir give proof soon...

    • @jbstatistics
      @jbstatistics  7 лет назад

      I'm not going to put up a formal proof anytime soon. A binomial random variable with parameters n and p can be thought of as the sum of n independent Bernoulli random variables with probability of success p. By the central limit theorem, this sum will be approximately normal for large n.

  • @jamesfullwood7788
    @jamesfullwood7788 4 года назад

    How does lambda stay constant as n tends to infinity?

    • @jbstatistics
      @jbstatistics  4 года назад

      n tends to infinity, p tends to 0, in such a fashion that np stays constant.

    • @jamesfullwood7788
      @jamesfullwood7788 4 года назад

      @@jbstatistics Ok so p is then a function of n, namely p(n)=\lambda/n

  • @umutaydn6184
    @umutaydn6184 10 месяцев назад

    thats great!

  • @venancejulius9938
    @venancejulius9938 3 года назад

    You didn't show how n raised by x canceled out

    • @jbstatistics
      @jbstatistics  3 года назад

      That's what 2:22- 3:00 is all about.

  • @capjus
    @capjus 10 лет назад +1

    thanks!! :)

  • @MrSazid1
    @MrSazid1 Год назад

    i need to learn to say 'Poisson' like him

  • @adamshenk9970
    @adamshenk9970 4 месяца назад

    bless wow thank yo

  • @NXT_LVL_DVL
    @NXT_LVL_DVL 3 года назад

    ❤️❤️❤️❤️❤️

  • @alex-my8hp
    @alex-my8hp 4 года назад

    i think the (x + 1)/n should be (x - 1)/n. not that it matters as it disappears in the limit

  • @jeffreya.faulkner8367
    @jeffreya.faulkner8367 3 года назад

    You have written n•(n-1) . . . (n-x+1) as n•(n-1) . . . (n-(x+1)) when it should be n•(n-1) . . . (n-(x-1)).