Produces Jackknife estimates and standard errors (R.W. Payne).
|Controls printed output (
||Data vectors from which the statistics are to be calculated|
||Other relevant information needed to calculate the statistics|
||Saves the variance-covariance matrix for the statistics|
||Texts, each containing a single line, to label the statistics|
||Saves the Jackknife estimate for each statistic|
||Saves Jackknife estimates of the standard errors|
||Saves the Jackknife pseudo-values|
||Saves the acceleration parameter for bias-corrected and accelerated bootstrap confidence intervals|
The Jackknife provides a way of decreasing bias and obtaining standard errors in situations where the standard methods might be expected to be inappropriate. The basic form of the Jackknife method works by calculating the statistic (or statistics) of interest omitting each data value in turn. Thus, if there are n data values, n “partial estimates” T-1 … T–n are obtained (where T–j is the estimate omitting value j). These are combined with the estimate T obtained from all the data, to produce n pseudo-values:
Pj = n × T – (n – 1) × T–j : j = 1 … n
The Jackknife estimate of the statistic is given by the mean of the pseudo-values, and the standard error by the standard error of the mean of the pseudo-values.
The Jackknife can be shown to eliminate the term proportional to 1/n from a bias of the form
T = t + a/n + O(1/n2)
where t is the true value of the estimate and O(1/n2) is a term of order one divided by the square of the number of observations (Quenouille 1956). However, it is not appropriate in all situations. In particular the statistic needs to be “smooth” (small changes in the data set should cause only small changes in the statistic); it will not work for example with medians or order statistics. Further details and advice are given by Miller (1974), Bissell & Ferguson (1975), Hinkley (1983) and Efron & Tibshirani (1993).
The data for
JACKKNIFE are provided as a list of vectors (variates, factors or texts) using the
DATA option. From this, new vectors are formed omitting each unit of the original vectors in turn, and a subsidiary procedure
RESAMPLE is called to calculate the statistics. Other relevant information can be provided for passing to
RESAMPLE, in any type of data structure, using the
ANCILLARY option. To use
JACKKNIFE, you need to provide a version of
RESAMPLE to calculate the particular statistics that you require. The default
RESAMPLE procedure, which accompanies
JACKKNIFE in the library, merely prints details of the syntax (also described in the Methods Section).
A label should be provided for each statistic, using the
LABEL parameter; by default, there is assumed to be a single statistic labelled simply as
Statistic. The estimates, their standard errors and variates of corresponding pseudo-values for each statistic can be saved by the
PSEUDOVALUES parameters, respectively. Also, if there is more than one statistic, a variance-covariance matrix can be saved for the estimates using the
Printed output is controlled by the
estimates for the estimates and their standard errors, and
vcovariance for the variance-covariance matrix; by default
The jackknife is also required for the calculation of bias-corrected and accelerated confidence limits for bootstrap statistics (as given by the
BOOTSTRAP procedure). The necessary acceleration quantities can be saved using the
ACCELERATION parameter. For details see Efron & Tibshirani, 1993, Section 14.3.
The original papers describing the Jackknife technique are by Quenouille (1949, 1956) and by Tukey (1958). Good expository accounts are provided by Hinkley (1983) or Bissell & Ferguson (1975).
JACKKNIFE needs a subsidiary procedure
RESAMPLE to calculate the statistics of interest.
RESAMPLE has an option,
DATA, which is used to supply the data vectors (variates, factors or texts) from which the statistics are to be calculated. (On the first occasion that
RESAMPLE is called, these will be the original vectors as supplied to
JACKNIFE, in order to calculate the estimate T; subsequently, they will be new vectors containing all except one of the units.) Other relevant information can can be supplied through the
ANCILLARY option, which corresponds to the
ANCILLARY option of
RESAMPLE can be called by the
BOOTSTRAP procedure, and it then also has an
AUXILIARY option, but this is not relevant to
There are two parameters:
STATISTICS supplies a list of scalars to store the estimates of each statistic, and
EXIT a list of scalars which should be set to zero or one according to whether or not each statistic could be estimated successfully with the supplied data vectors. If the value of
EXIT is not calculated in
JACKKNIFE assumes that the calculations succeeded. This example shows a version of
RESAMPLE which calculates the correlation between two variates.
PROCEDURE [PARAMETER=pointer] 'RESAMPLE'
OPTION 'DATA', " (I: variates, factors or texts) data
vectors from which to calculate the
statistics; no default"\
'ANCILLARY'; " (I: any type of structure) other
relevant information needed to
calculate the statistics "\
SET=yes,no; LIST=yes; DECLARED=yes; PRESENT=yes
PARAMETER 'STATISTIC', " (O: scalars) to save the calculated
'EXIT'; " (O: scalars) to save an exit code
to indicate failure (EXIT[i]=1) or
success (EXIT[i]=0) when calculating
MODE=p; TYPE='scalar'; SET=yes
CALCULATE STATISTIC = CORRELATION(DATA; DATA)
& EXIT = STATISTIC==C('missing')
If any of the data vectors is restricted,
JACKKNIFE will use only the units that are not restricted for any of the vectors.
Bissell, A.F. & Ferguson, R.A. (1975). The jackknife – toy, tool or two-edged weapon. The Statistician, 24, 79-100.
Efron, B. & Tibshirani, R.J. (1993). An Introduction to the Bootstrap. Chapman & Hall, London.
Hinkley, D. (1983). Jackknife methods. In: Encyclopedia of Statistics, Volume 4 (ed: S. Kotz, N.L. Johnson & C.B. Read). Wiley, New York.
Miller, R.G. (1974). The jackknife – a review. Biometrika, 61, 1-15.
Quenouille, M.H. (1949). Approximate tests of correlation in time series. Journal of the Royal Statistical Society, Series B, 11, 18-44.
Quenouille, M.H. (1956). Notes on bias in estimation. Biometrika, 61, 353-360.
CAPTION 'JACKKNIFE example',!t(\ 'The data are scores from two tests on new admissions to Law School',\ '(Efron, 1981, The Jackknife, the Bootstrap & Other Resampling',\ ' Plans. CBMS Monograph 38, SIAM, Philadelphia); listed in Table 1',\ 'of Hinkley (1983, Encyclopedia of Statistics Volume 4, page 282).');\ STYLE=meta,plain " Define RESAMPLE to calculate the correlation between the two scores." PROCEDURE [PARAMETER=pointer] 'RESAMPLE' OPTION 'DATA', " (I: variates, factors or texts) data vectors from which to calculate the statistics; no default"\ 'AUXILIARY', " (I: pointers) auxiliary sets of data vectors, each of which is to be resampled independently"\ 'ANCILLARY'; " (I: any type of structure) other relevant information needed to calculate the statistics "\ MODE=p; TYPE=!t(variate,factor,text),'pointer',*; SET=yes,no,no;\ LIST=yes; DECLARED=yes; PRESENT=yes PARAMETER 'STATISTIC', " (O: scalars) to save the calculated statistics "\ 'EXIT'; " (O: scalars) to save an exit code to indicate failure (EXIT[i]=1) or success (EXIT[i]=0) when calculating each STATISTIC[i]"\ MODE=p; TYPE='scalar'; SET=yes CALCULATE STATISTIC = CORRELATION(DATA; DATA) & EXIT = STATISTIC==C('missing') ENDPROCEDURE VARIATE [VALUES=576,635,558,578,666,580,555,661,651,605,653,575,545,572,594] Y & [VALUES=3.39,3.30,2.81,3.03,3.44,3.07,3.00,3.43,3.36,3.13,\ 3.12,2.74,2.76,2.88,2.96] Z JACKKNIFE [DATA=Y,Z] 'Correlation'