SISSI: a Resampling-Based Software for Sample Size Determination

The determination of sample size before collecting experimental data is a key issue to obtain reliable measurements of data which are representative of the system under study. Software tools do exist to support researchers in sample size determination, but they largely rely on statistical assumptions of data distribution, which are not likely to occur when working with biological samples. In particular, conventional inferential methods based on Student-t distribution are inappropriate under non-normal data distribution. Statistical procedures known as resampling methods (Efron and Tibshirani, 1991) make intensive use of the information contained in one sample only and are suitable to skip over limitations associated to parametric statistics. A particular resampling method, derived from the jackknife (Hinkley, 1983) and recognized as visual jackknife, was recently proposed by Confalonieri (2004) and applied by Confalonieri et al. (2006). According to this approach, virtual samples are generated from the N data of a sample, as reduced bodies of data leaving out a group of k elements at each time. A user defined number of virtual samples (an adequate number is typically within 250 and 500) is generated for all k values in the range from k=1 to k=N-2. Optimal sample size is identified with the size N-k giving asymptotic stabilization of sample mean, i.e. its variability does not really decrease with further sample size increase. Software tool SISSI (Shortcut In Sample Size Identification) implements the visual jackknife approach for use in sample size determination and is illustrated here.

BELLOCCHI Gianni;  CONFALONIERI Roberto;  ACUTIS Marco;  GENOVESE Giampiero; 

2010-09-28

European Society for Agronomy (ESA)

JRC33940

http://www.esa.org.pl/,    https://publications.jrc.ec.europa.eu/repository/handle/JRC33940,