Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Conditional probability for normally R.V

  1. Jan 14, 2014 #1

    w1,w2,w3,w4,w5 are normally distributed i.i.d R.V.

    I want to calculate the probability that slots 1&2(w1,w2) are smaller than slots 3,4,5 (where mu>0)

    I'm able to calculate the probability that slot 1 is smaller than 3,4,5, now i'm stuck on the calculation for slot 2, where i know that i have to consider the outcomes from the first calculation, for example if i know that w1(slot 1) is larger than w2(slot 2), than the probability that w2(slot 2) is smaller than (3,4,5) is 1.

    until now i was sure that the whole probability is :

    P_{correct}= P(w_1<\mu+w_3)P(w_1<\mu+w_4)P(w_1<\mu+w_5)P(w_2<\mu+w_3)P(w_2<\mu+w_4)P(w_2<\mu+w_5)
    now i am thinking of:
    P_{correct}= P(w_1<\mu+w_3)P(w_1<\mu+w_4)P(w_1<\mu+w_5)P(w_2<\mu+w_3|w_1<w_2)P(w_2<\mu+w_4|w_1<w_2)P(w_2<\mu+w_5|w_1<w_2)P(w_2<\mu+w_3|w_2<w_1)P(w_2<\mu+w_4|w_2<w_1)P(w_2<\mu+w_5|w_2<w_1)
    $$where the last 3 terms are equal to 1 because i have the knowledge that $$P(w1<μ+w3)P(w1<μ+w4)P(w1<μ+w5) $$
    any help will be appreciated.
  2. jcsd
  3. Jan 15, 2014 #2
    How do i edit my post ? i want to revise the question
  4. Jan 16, 2014 #3
    rewrite my question, i didn't find how to just edit

    the following modulation is in the frequency domain, x-axis is frequency , y-axis is amplitude of electric current.

    the signal is a pulse divided to 5 equally slots by means of electrical current where 2 "holes" indicating the binary message.
    for this example my constellation size is 2^3. therefore 5C2 = 10 combinations is possible in this setup (5 slots and 2 "holes") and i'll will use 8 of them.

    i am trying to calculate probability as described :


    w1,w2,w3,w4,w5 are normally distributed i.i.d R.V (AWGN).

    A correct detection is when the noise in slots 1 and 2 both is smaller than the noise + mu(the current related to the slot).


    so the probability for correct detection as i'm thinking is:

    1. P_{correct}= P(w_1<\mu+w_3\cap w_1<\mu+w_4\cap w_1<\mu+w_5\cap w_2<\mu+w_3\cap w_2<\mu+w_4\cap w_2<\mu+w_5)

    if i'm right, i can calculate just one of the terms , because of the i.i.d properties.
    2. P_{correct}= P(w_1<\mu+w_3)P(w_1<\mu+w_4)P(w_1<\mu+w_5)P(w_2<\mu+w_3)P(w_2<\mu+w_4)P( w_2<\mu+w_5)
    P_{correct}= P(w_1<\mu+w_3)^6
    P(w_1<\mu+w_3)= \int_{-\infty}^{\infty}\frac{1}{\sqrt{2\pi}\sigma}e^{-\frac{w_1^2}{2\sigma^2}}Q(\frac{w_1-\mu}{\sigma})dw_1
    hence the error probability is :
    therefore the general case with n- slots and k- "holes" the probability is
    i did another calculation with my friend and we got:

    i'm trying to understand which of Eq.3 or Eq.4 is the correct answer ? and the transition from Eq.1 to Eq.2 is legal, meaning, are slots 1,2,3,4,5 are absolutely independent ?
  5. Jan 16, 2014 #4

    Stephen Tashi

    User Avatar
    Science Advisor

    A way to put that in familiar territory is:

    [itex]max(w_1,w_2) < min(\mu+w_2,\mu+w_4,\mu+w_5) [/itex]

    Look at the theory of "order statistics" to find out how the max and min are distributed.
  6. Jan 19, 2014 #5
    thanks i'm hoping it will do the work.
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook