+ Reply to Thread
Results 1 to 8 of 8

Thread: Joint Normal CDF

  1. #1
    Points: 3,089, Level: 34
    Level completed: 26%, Points required for next Level: 111

    Posts
    4
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Joint Normal CDF




    Hello

    I am new to this forum as well as to actuary field. I have this question:

    There are two random variables X_1 and X_2 following standard normal distributions:
    X_1 ~ N(0,1)
    X_2 ~ N(0,1)
    If we now set
    Y_1 = X_1
    Y_2 = X_1+X_2
    then we know that
    Y_1 ~ N(0,1)
    Y_2 ~ N(0,2)
    The question is how to find the joint CDF Of Y = (Y_1, Y_2). Can we find the joint CDF in a closed form? If yes, how?

    This is a problem arising in the field of risk claims.

    Is there anyone who can guide me in resolving this issue?

    Many thanks in advance

    Kind regards
    hpriye

  2. #2
    Dark Knight
    Points: 6,762, Level: 54
    Level completed: 6%, Points required for next Level: 188
    vinux's Avatar
    Posts
    2,011
    Thanks
    52
    Thanked 241 Times in 205 Posts

    Question

    If you know the joint CDF Of (X_1, X_2) then we can find the joint CDF of Y. Or are you assuming X_1 and X_2 are independent?

    Vinux

  3. #3
    Points: 3,089, Level: 34
    Level completed: 26%, Points required for next Level: 111

    Posts
    4
    Thanks
    0
    Thanked 0 Times in 0 Posts
    Many thanks!

    Yes, indeed X_1 and X_2 are independent standard normal.

    So what is the procedure for finding the joint CDF of Y=(Y_1, Y_2), where Y_1 = X_1 and Y_2 = X_1+X_2?

    Kind regards
    hpriye

  4. #4
    Dark Knight
    Points: 6,762, Level: 54
    Level completed: 6%, Points required for next Level: 188
    vinux's Avatar
    Posts
    2,011
    Thanks
    52
    Thanked 241 Times in 205 Posts

    Exclamation

    Then it is not difficult.
    In this case Y follows BVN(bivariate normal distribtion) with mean [ 0 0]' and Variance matrix
    V(Y) = [ 1 1; 1 2]
    ( details: V(Y_1) = V(X_1) = 1 , V(Y_2) = V(X_1) + V(X_2) = 2
    & Cov(Y_1,Y_2) = E{X_1*(X_1+X_2)} = E(X_1*X_1) + E(X_1)*E(X_2) = 1+0 =1 )

    I think this solve the problem
    ------------------------------------
    And if you wanted to calculate CDF at particular point EG: Y_1 = 1 and Y_2 = 4
    i e X_1 = 1 & X2_1 = 4-1 = 3

    so CDF Y(1,4) = CDF X(1,3) = CDF(1) * CDF(3)

    Regards
    Vinux aka Richie

  5. #5
    Points: 3,089, Level: 34
    Level completed: 26%, Points required for next Level: 111

    Posts
    4
    Thanks
    0
    Thanked 0 Times in 0 Posts
    Hello

    Whatever you have explained is for finding the joint probability distribution function (PDF) and not the cumulative distribution function (CDF)

    The definition of CDF is (in the case of normal distribution)

    doubleIntegral (PDF function, y_2=-infinity to v, y_1=-infinity to u);

    ... and what I am looking for is a closed form expression for the above integral which is perhaps impossible - as I have now realized.

    .. But thanks for your efforts and time

    Kind regards
    hpriye

  6. #6
    Dark Knight
    Points: 6,762, Level: 54
    Level completed: 6%, Points required for next Level: 188
    vinux's Avatar
    Posts
    2,011
    Thanks
    52
    Thanked 241 Times in 205 Posts

    Wink

    The double integreal part can be replaced by product of two single integral.Since X_1 and X_2 are independent
    and Single integral ( Univariate cumilative normal probabilities ) are avaliable in most of the tools ( Excel Sas R etc)
    Details:
    If you wanted to convert CDF of Y then
    First convert Y in terms of X then
    CDF Y(1,4) = CDF X(1,3) = CDF(1) * CDF(3) ( see the example mentioned above)
    Here CDF(1) = integral ( -inf to 1) std normal density) =~ 0.84
    and CDF(3) =~ 0.99
    so CDF Y(1,4) = 0.84*.99 = 0.83

    Regards
    Vinux aka Richie

  7. #7
    Points: 3,089, Level: 34
    Level completed: 26%, Points required for next Level: 111

    Posts
    4
    Thanks
    0
    Thanked 0 Times in 0 Posts
    Hello

    No, it is not possible to write it as product of two integrals for the PDF is :

    1/(2*pi) exp(-1/2*y_1^2)*exp(-1/2*(y_2-y_1)^2)

    to be integrated from -infinity to v and -infinity to u

    Kind regards
    hpriye

  8. #8
    Dark Knight
    Points: 6,762, Level: 54
    Level completed: 6%, Points required for next Level: 188
    vinux's Avatar
    Posts
    2,011
    Thanks
    52
    Thanked 241 Times in 205 Posts

    Quote Originally Posted by hpriye View Post
    Hello

    No, it is not possible to write it as product of two integrals for the PDF is :

    1/(2*pi) exp(-1/2*y_1^2)*exp(-1/2*(y_2-y_1)^2)

    to be integrated from -infinity to v and -infinity to u

    Kind regards
    hpriye
    ---------------------------
    Yup.. you are true and thank you.. CDF of Y will be in the form of double integral. I think it can't be simplify further.

+ Reply to Thread

           




Similar Threads

  1. Normal and Standard normal distributed variable
    By johny_5 in forum Statistics
    Replies: 7
    Last Post: 01-20-2011, 04:56 AM
  2. Joint Independent Normal Random Variable question
    By StatStudent3 in forum Probability
    Replies: 3
    Last Post: 11-11-2010, 09:57 PM
  3. Joint distributions
    By Cookie in forum Statistics
    Replies: 1
    Last Post: 05-26-2010, 08:51 AM
  4. Joint Pdf
    By Gp7417 in forum Probability
    Replies: 1
    Last Post: 12-20-2009, 11:18 AM
  5. Joint reference prior for normal distribution
    By russell in forum Psychology Statistics
    Replies: 0
    Last Post: 03-27-2006, 09:35 PM

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts






Advertise on Talk Stats