Presentation is loading. Please wait.

Presentation is loading. Please wait.

Point processes on the line. Nerve firing.

Similar presentations


Presentation on theme: "Point processes on the line. Nerve firing."— Presentation transcript:

1 Point processes on the line. Nerve firing.

2 Stochastic point process. Building blocks
Process on R {N(t)}, t in R, with consistent set of distributions Pr{N(I1)=k1 ,..., N(In)=kn } k1 ,...,kn integers  0 I's Borel sets of R. Consistentency example. If I1 , I2 disjoint Pr{N(I1)= k1 , N(I2)=k2 , N(I1 or I2)=k3 } =1 if k1 + k2 =k3 = 0 otherwise Guttorp book, Chapter 5

3 Points: ... -1  0  1  ... discontinuities of {N} N(t) = #{0 < j  t} Simple: j  k if j  k points are isolated dN(t) = 0 or 1 Surprise. A simple point process is determined by its void probabilities Pr{N(I) = 0} I compact

4 Conditional intensity. Simple case
History Ht = {j  t} Pr{dN(t)=1 | Ht } = (t:)dt  r.v. Has all the information Probability points in [0,T) are t1 ,...,tN Pr{dN(t1)=1,..., dN(tN)=1} = (t1)...(tN)exp{- (t)dt}dt1 ... dtN [1-(h)h][1-(2h)h] ... (t1)(t2) ...

5 Parameters. Suppose points are isolated
dN(t) = if point in (t,t+dt] = otherwise 1. (Mean) rate/intensity. E{dN(t)} = pN(t)dt = Pr{dN(t) = 1} j g(j) =  g(s)dN(s) E{j g(j)} =  g(s)pN(s)ds Trend: pN(t) = exp{+t} Cycle:  + cos(t+)  0

6 Product density of order 2.
Pr{dN(s)=1 and dN(t)=1} = E{dN(s)dN(t)} = [(s-t)pN(t) + pNN (s,t)]dsdt Factorial moment

7 Autointensity. Pr{dN(t)=1|dN(s)=1} = (pNN (s,t)/pN (s))dt s  t = hNN(s,t)dt = pN (t)dt if increments uncorrelated

8 Covariance density/cumulant density of order 2.
cov{dN(s),dN(t)} = qNN(s,t)dsdt st = [(s-t)pN(s)+qNN(s,t)]dsdt generally qNN(s,t) = pNN(s,t) - pN(s) pN(t) st

9 Identities. 1. j,k g(j ,k ) =  g(s,t)dN(s)dN(t) Expected value. E{ g(s,t)dN(s)dN(t)} =  g(s,t)[(s-t)pN(t)+pNN (s,t)]dsdt =  g(t,t)pN(t)dt +  g(s,t)pNN(s,t)dsdt

10 2. cov{ g(j ),  h(k )} = cov{ g(s)dN(s),  h(t)dN(t)} =  g(s) h(t)[(s-t)pN(s)+qNN(s,t)]dsdt =  g(t)h(t)pN(t)dt +  g(s)h(t)qNN(s,t)dsdt

11 Product density of order k.
t1,...,tk all distinct Prob{dN(t1)=1,...,dN(tk)=1} =E{dN(t1)...dN(tk)} = pN...N (t1,...,tk)dt1 ...dtk

12

13 Proof of Central Limit Theorem via cumulants in i.i.d. case.
Normal distribution facts. 1. Determined by its moments 2. Cumulants of order  2 identically 0 Y1, Y2, ... i.i.d. mean 0, variance 2, all moments, E{Yk} k=1,2,3,4,... existing Sn = Y1 + Y Yn E{Sn } = var{ Sn} = n 2 cumr Sn = n r cumr Y = cum{Y,...,Y} cumr {Sn / n} = n r / nr/2  0 for r = 3, 4, ...  2 r = as n  

14 Cumulant density of order k.
t1,...,tk distinct cum{dN(t1),...,dN(tk)} = qN...N (t1 ,...,tk)dt1 ...dtk

15 Stationarity. Joint distributions, Pr{N(I1+t)=k1 ,..., N(In+t)=kn} k1 ,...,kn integers  0 do not depend on t for n=1,2,... Rate. E{dN(t)=pNdt Product density of order 2. Pr{dN(t+u)=1 and dN(t)=1} = [(u)pN + pNN (u)]dtdu

16 Autointensity. Pr{dN(t+u)=1|dN(t)=1} = (pNN (u)/pN)du u  0 = hN(u)du Covariance density. cov{dN(t+u),dN(t)} = [(u)pN + qNN (u)]dtdu

17

18

19

20 Taking "E" again,

21 Association. Measuring? Due to chance?
Are two processes associated? Eg. t.s. and p.p. How strongly? Can one predict one from the other? Some characteristics of dependence: E(XY)  E(X) E(Y) E(Y|X) = g(X) X = g (), Y = h(),  r.v. f (x,y)  f (x) f(y) corr(X,Y)  0

22 Bivariate point process case.
Two types of points (j ,k) Crossintensity. Prob{dN(t)=1|dM(s)=1} =(pMN(t,s)/pM(s))dt Cross-covariance density. cov{dM(s),dN(t)} = qMN(s,t)dsdt no ()

23

24 Mixing. cov{dN(t+u),dN(t)} small for large |u| |pNN(u) - pNpN| small for large |u| hNN(u) = pNN(u)/pN ~ pN for large |u|  |qNN(u)|du <  See preceding examples

25 Power spectral density. frequency-side, , vs. time-side, t
/2 : frequency (cycles/unit time) Non-negative Unifies analyses of processes of widely varying types

26 Examples.

27

28 Spectral representation. stationary increments - Kolmogorov

29 Algebra/calculus of point processes.
Consider process {j, j+u}. Stationary case dN(t) = dM(t) + dM(t+u) Taking "E", pNdt = pMdt+ pMdt pN = 2 pM

30 Taking "E" again,


Download ppt "Point processes on the line. Nerve firing."

Similar presentations


Ads by Google