Nonparametric Statistical Methods - Stony Brook University资料.ppt

* * Jackknife Method1First implemented by R.E. von Mises2 (1883-1953), then developed (separately) by  Tukey (1915-2000) and Quenouille in the 1950s. Resamples by deleting one observation at a time This method is also useful for estimating the standard error of a statistic, say based on a random sample of size n drawn from some distribution ‘F’. First, calculate the n values of the statistic denoted by Let and be the standard deviation of The jackknife estimate of is given by * 14.6 Resampling Methods SAS can be used to perform permutation, bootstrap, and jackknife resampling. For the most part macros are required. These can be written and are also readily available on the web. PROC MULTTEST can be used to perform several tests incorporating permutation or bootstrap In the following two examples, We use permutation and bootstrap resampling to obtain t-test p-value adjustment. * 14.6 Resampling Methods Example 14.15 permutation test and data capacitor; Input group $ failtime @@; Datalines; Control 17.9 control 23.7 control 29.8 Stressed 15.2 stressed 18.3 stressed 21.1 ; Proc multtest data=capacitor permutation nsample=25000 out=results outsamp=samp; test mean(failtime /lower); class group; contrast a vs b -1 1; Run; proc print data=samp(obs=18); run; proc print data=results; run; The PERMUTATION option in the PROC MULTTEST statement requests permutation resampling, and NSAMPLE=25 000 requests 25000 permutation samples. The OUTSAMP=SAMP option creates an output SAS data set containing the 25000 permutation samples. The TEST statement specifies the t-test for T. The test is lower-tailed. The grouping variable in the CLASS statement is group, and the coefficients across the groups are -1 and 1, as specified in the CONTRAST statement. (See Chapter 12) PROC PRINT displays the first 18 observations of the Res data set containing the bootstrap samples. * 14.6 Resampling Methods * 14.6 Resampling Methods Example 14.17 b