home *** CD-ROM | disk | FTP | other *** search
- .MCD 20000 0
- .CMD PLOTFORMAT logs=0,0 subdivs=1,1 size=5,15 type=l
- .CMD FORMAT rd=d ct=10 im=i et=3 zt=15 pr=3 mass length time charge
- .CMD SET ORIGIN 0
- .CMD SET TOL 0.001000
- .CMD MARGIN 0
- .CMD LINELENGTH 78
- .CMD SET PRNCOLWIDTH 8
- .CMD SET PRNPRECISION 4
- .TXT 0 40 1 38
- a1,37,38,36
- Copyright (c) 1988 by MathSoft, Inc
- .TXT 2 -11 1 22
- a1,21,77,20
- JACKKNIFE ESTIMATES
- .TXT 2 -28 6 80
- a6,79,77,443
- The jackknife process provides a way of estimating the standard error of a
- statistic calculated from a sample. To jackknife the data, you divide your
- sample into k groups and recalculate the statistic in question k times,
- removing the effect of each one of the k groups in turn. The mean of these
- k pseudo-values is an estimate for the statistic, and their standard
- deviation gives an estimate for the standard error of this estimate.
- .TXT 6 0 3 80
- a3,79,77,232
- Suppose that the sample is stored in the array V and that the sample function
- to be calculated is f(V). Here V contains 10 exponential deviates and f is
- the mean absolute deviation. The groups are the individual elements of V.
- .EQN 3 42 2 30
- f(V)~mean((|(V-mean(V))){49})
- .EQN 1 -42 1 12
- j~0;20
- .EQN 0 19 2 18
- V[j~-ln(rnd(1))
- .TXT 2 -19 1 68
- a1,67,77,66
- Now define a family of samples each leaving out one element of V:
- .EQN 2 0 1 15
- k~length(V)
- .EQN 0 19 1 15
- i~0;k-2
- .EQN 0 18 1 15
- j~0;k-1
- .EQN 2 -37 2 19
- J[(i,j)~V[(i+(i≥j))
- .TXT 0 22 1 51
- a1,50,55,49
- Column j of J is V with the jth element omitted.
- .TXT 3 -23 1 15
- a1,14,78,13
- [Ctrl][PgDn]
- .TXT 2 1 1 76
- a1,75,77,74
- The jth pseudo-value is calculated using the sample with group j omitted:
- .EQN 2 0 3 31
- P[j~k*f(V)-(k-1)*f(J{52}j)
- .TXT 4 0 1 36
- a1,35,78,34
- The jackknifed value for f(V) is:
- .EQN 0 48 1 19
- mean(P)=?
- .EQN 2 0 4 19
- \(var(P)/(k-1))=?
- .TXT 2 -48 1 41
- a1,40,78,39
- An estimate for its standard error is:
-