ABSTRACT
Most of the inferential procedures that we have studied were devised before the advent of computers. Student’s 1-sample t-test was introduced in 1908! Before computers, statistical inference relied on procedures for which (1) the quantities that depend on the data can be computed by hand, and (2) the sampling distributions needed to compute critical values are sufficiently tractable that the critical values can be tabled for widespread use. For decades, data analysis required tedious calculation and access to tables of quantiles for the standard normal and various chi-squared, t, and F distributions. Although the enormous practical value of procedures such as t-tests, ANOVA, and simple linear regression is beyond dispute, one reason that these procedures were central to 20th-century statistics was the simple fact that it was possible for a user to perform the necessary calculations. As computing power increased, the practice of statistics began to change. First desktop calculators, then computers, were used to ease the computational burden of the same familiar procedures. Eventually, computationally intensive procedures were introduced that would have been unthinkable in the days of Student and Fisher.
