• yetAnotherUser@discuss.tchncs.de
    link
    fedilink
    English
    arrow-up
    4
    ·
    edit-2
    22 hours ago

    I’ve found a proper approximation after some time and some searching.

    Since the binomial distribution has a very large n, we can use the central limit theorem and treat it as a normal distribution. The mean would be obviously 500 billion, the standard deviation is √(n * p * (1-p)) which results in 500,000.

    You still cannot plug that into WA unfortunately so we have to use a workaround.

    You would calculate it manually through:

    Φ(b) - Φ(a), with
    b = (510 billion - mean) / (standard deviation) = 20,000
    and
    a = (490 billion - mean) / (standard deviation) = -20,000
    and
    Φ(x) = 0.5 * (1 + erf(x/√2))
    

    erf(x) is the error function which has the neat property: erf(-x) = -erf(x)

    You could replace erf(x) with an integral but this would be illegible without LaTeX.

    Therefore:

    Φ(20,000) - Φ(-20,000)
    = 0.5 * [ erf(20,000/√2) - erf(-20,000/2) ]
    = erf(20,000/√2)
     erf(14,142)
    

    WolframAlpha will unfortunately not calculate this either.

    However, according to Wikipedia an approximation exists which shows that:

    1 - erf(x) ≈ [(1 - e^(-Ax))e^(-x²)] / (Bx√π)
    

    And apparently A = 1.98 and B = 1.135 give good approximations for all x≥0.

    After failing to get a proper approximation from WA again and having to calculate every part by itself, the result is very roughly around 1 - 10^(-86,857,234).

    So it is very safe to assume you will lose between 49% and 51% of your gut bacteria. For a more realistic 10 trillion you should replace a and b above with around ±63,200 but I don’t want to bother calculating the rest and having WolframAlpha tell me my intermediary steps are equal to zero.

    • enkers@sh.itjust.works
      link
      fedilink
      English
      arrow-up
      2
      ·
      21 hours ago

      Whoa, good work! I think I’m going to have to go over this a few times to grock how it works, especially the Φ(b) - Φ(a) bit. My stats textbook has a bit too much dust on it. ;)