Examples (BOOTSTRAP command)

Simple Resampling; Maintaining a Consistent Case Basis

BOOTSTRAP.
DESCRIPTIVES VARIABLES=var1 var2 var3
  /MISSING=VARIABLE.
  • The BOOTSTRAP command requests 1000 bootstrap samples.
  • No variables are specified on the BOOTSTRAP command, so no records are deleted from the resampling. This allows the DESCRIPTIVES procedure to use variablewise deletion of missing values on the full set of records; however, the case basis will be inconsistent across bootstrap resamples, and inferences made from the results would be questionable. 
    BOOTSTRAP
  /VARIABLES ANALYSIS(INPUT=var1 var2 var3).
DESCRIPTIVES VARIABLES=var1 var2 var3
  /STATISTICS MEAN STDDEV MIN MAX
  /MISSING=VARIABLE.
  • This is the same as the previous analysis, but variables var1, var2, and var3 are used to determine the case basis for resampling. Records with missing values on any of these variables are deleted from the analysis.
  • The DESCRIPTIVES procedure following BOOTSTRAP is run on the bootstrap samples.
  • The STATISTICS subcommand produces the mean, standard deviation, minimum, and maximum for variables var1, var2, and var3 on the original data. Additionally, pooled statistics are produced for the mean and standard deviation.
  • Even though the MISSING subcommand specifies variablewise deletion of missing values, the listwise deletion performed by BOOTSTRAP is what determines the case basis. In effect, the MISSING specification on DESCRIPTIVES is irrelevant here.

Stratified Resampling

BOOTSTRAP
  /VARIABLES SAMPLING(STRATA=strataVar)
             ANALYSIS(INPUTS=var1).
DESCRIPTIVES var1.
  • The BOOTSTRAP command requests 1000 bootstrap samples stratified by strataVar.
  • Variables var1 and strataVar are used to determine the case basis for resampling. Records with missing values on these variables are deleted from the analysis.
  • The DESCRIPTIVES procedure following BOOTSTRAP is run on the bootstrap samples, and produces the mean, standard deviation, minimum, and maximum for the variable var1 on the original data. Additionally, pooled statistics are produced for the mean and standard deviation.