XLSTAT - Resampling
Principles of resampling
Resampling methods have become more and more popular since computational power has increased. It is a well-known approach to nonparametric statistics. The principle is very simple: from your original sample, randomly draw a new sample and recalculate statistics. Repeating this step many times gives you the empirical distribution of the statistic, from which you obtain the standard error, and confidence intervals.
Resampling in XLSTAT
With XLSTAT, you can apply these methods on a selected number of descriptive statistics for quantitative data.
Three resampling methods are available:
It is the most famous approach; it has been introduced by Efron and Tibisharni (1993). It is a statistical method for estimating the sampling distribution of an estimator by sampling with replacement from the original sample. The number of samples has to be given.
- Random without replacement:
Subsamples are drawn randomly from the original sample. The size of the subsample has to be specified.
The sampling procedure is based on suppressing one observation to the original sample (of size N). Each subsample has N-1 observations and the process is repeated N times. It is less robust than the bootstrap.
Although you can select several variables (or samples) at the same time, XLSTAT calculates all the descriptive statistics for each of the samples independently.
This analysis is available in the XLStat-Basic addin for Microsoft Excel™