Kolmogorov Forward differential equation

Discussion in 'CT4' started by Indy007, Mar 21, 2009.

  1. Indy007

    Indy007 Member

    Hi!

    I'm looking at Chapter 5 section 4 of the notes (page 20), and am having difficulty following the logic behind the calculation of the Kolmogorov Forward equation.

    Specifically, I'm having problem with the first step where:

    Sum[Pik(t)Pkj(h)] = Pij(t) +h*Sum[Pik(t) *MUkj +o(h)]

    I can't seem to figure out how this has been determined. I've tried using the relationship as specified on page 20, but end up with the following:

    Sum[Pik(t)Pkj(h)] = Pij(t)(1+MUjj+oh) +h*Sum[Pik(t) *MUkj +o(h)]

    Apologies if I've overlooked something obvious, but have been staring at the notes for a while now and can't seem to wrap my head round it!:confused:

    I'd appreciate any help you can give me.

    ~Indy
     
  2. jensen

    jensen Member

    My guess is, you should not substitute them directly, instead write it like this:

    = Sum P_ik(t) { P_kj(h) + h.Mu_kj + o(h)}

    where if k = j, then the {} part becomes 1 + h.Mu_jj + o(h)
    and if k =/ j, then {} becomes h.Mu_kj + (o)h
    just like how 5.7 is.

    Then, breaking up the summation:

    = Sum_k=j P_ik(t).P_kj(h) + ... (the part you have correctly)
    but the first part is really P_ij(t), so there you have it.

    Can someone please confirm my derivation, please?

    Thanks.
     
  3. Michael

    Michael Member

    I think you're on the right track Indy, just need one more step.

    You've got

    Sum[Pik(t)Pkj(h)] = Pij(t)(1+MUjj+oh) +h*Sum [Pik(t) *MUkj +o(h)]
    = Pij(t) + Pij(t)(MUjj + oh) + h*Sum [Pik(t) *MUkj +o(h)]

    where the summation is for values of k not equal to j.
    By changing the summation to take all values of k, the Pij(t)(MUjj + oh) gets included as well, so you're left with

    Pij(t) +h*Sum[Pik(t) *MUkj +o(h)]

    as required.
     
  4. jensen

    jensen Member

    Hi Michael

    How do you combine the 2nd and 3rd term together when one has "h" and the other does not?
     
    Last edited by a moderator: Mar 23, 2009
  5. Michael

    Michael Member

    Hrm... Er, clearly you can't. :) Guess I should have looked a little more closely before posting...

    The equation that Indy posted:

    Sum[Pik(t)Pkj(h)] = Pij(t)(1+MUjj+oh) +h*Sum[Pik(t) *MUkj +o(h)]

    isn't quite correct, I think it should be

    Sum kES[Pik(t)Pkj(h)]
    = Pij(t)Pjj(h) + Sum k!=j[Pik(t)Pkj(h)]
    = Pij(t)[1+h*MUjj+o(h)] + Sum k!=j [Pik(t) [h*MUkj +o(h)]]

    Now you can take the Pij(t) * (hMUjj + o(h)) into the summation.
     
  6. CorkActuary

    CorkActuary Member

    For the FDE of P3,7(t), is it:

    [row 4 of P]*[column 8 of A]

    i.e. [P30(t),P31(t),...]*[Generator matrix values for column 8]
     

Share This Page