{\rtf1\ansi\ansicpg1252\deff0\nouicompat\deflang3079{\fonttbl{\f0\fnil\fcharset0 Arial;}{\f1\fnil\fcharset2 Symbol;}} {\*\generator Riched20 6.3.9600}\viewkind4\uc1 \pard\qc\b\fs28 Sample Size Influence View\b0\par \pard\fs18\par \par \pard\qj The Sample Size Influence View should help answer the question: "Have I done enough experiments/repetitions?". Because this is not so easy to answer and generalize, this view helps to get a better feeling for the data. It groups the samples into groups of different size and shows for each group a box plot. This can be done in two different ways:\par \par \pard{\pntext\f1\'B7\tab}{\*\pn\pnlvlblt\pnf1\pnindent0{\pntxtb\'B7}}\fi-360\li720\qj In the first case samples are randomly drawn from the data to fill the first group. The drawn samples are then put back and then it is again drawn to fill the second group and so on. Because the drawn samples are "put back" this is also called "drawing with replacement". \par {\pntext\f1\'B7\tab}If the "Hypergeometric" checkbox is checked, the randomly drawn samples are not put back ("drawing without replacement"). \par \pard\qj\par This should give an idea of how different sample sizes of the data look like, e.g. how many outliers, how the average or standard deviation changes with increased sample size. With the "Recalculate" button the drawing can be repeated. You can see how different group sizes fluctuate. This effect should get less with bigger group sizes. \par \par The recommended sample size shown in the upper right corner is calculated as described in \i G\'f6ran Kauermann, Helmut K\'fcchenhoff: Stichproben: Methoden und praktische Umsetzung mit R, section 2.27 \i0 with a confidence intervall of 0.95. Please note that this may be a very pessimistic estimation and that can often be considered as an upper bound for the sample size. \par \pard\par \par }