Measuring the variability of a statistical series
Statistical analysis is used where the thing to be studied can be reduced to or stated in terms of numbers. Not all the undertakings that rely on measurements ordinarily employ statistical analyses. In surveying, physics, and chemistry, for example, the particular thing being studied can usually be measured so closely, and varies over such a small range, that the true value can be established within narrow limits. But even in these fields, the modern work on atomic subparticles has involved the use of statistical concepts. In fact, the statistical concept of true value owes its existence to the reproducibility of measurements in fields like these.
In many natural sciences, the problem to be studied can be simplified by the use of controlled experimental conditions, which permit the influence of various factors to be studied one at a time. In such sciences, statistical methods can be used to plan experiments in such a way as to make the conclusions most reliable with a minimum of effort, and they can be used to measure the interrelations in sciences like astronomy, where the phenomena can be observed but not controlled.1
In the social sciences, there are fewer opportunities for the use of controlled experiments. Such sciences have to rely on statistical analysis, both to judge the importance of observed differences and to untangle the separate effects of multiple factors. Statistical analysis is used in the study of occurrences where the true value or relation cannot be measured____________________
R. A. Fisher, The Design of Experiments, 5th ed., Oliver and Boyd, Edinburgh and London, 1949.