A General Computer Programme for the Analysis of Factorial Experiments

Iterative Procedures for Missing Values in Experiments

“…Experience has shown that the same modification is usually advantageous for more than one missing value in a variate; with p as the initial value for each missing unit, the modification often reduces (and sometimes much reduces) the number of iterations needed. Accordingly, although the multiplier n/E will not necessarily give the fewest iterations, the following scheme (successfully adopted in a revised version of the Rothamsted General Factorial Program described by Yates and Anderson (1966)) is recommended for any new general program:…”

Computers as an Instrument for Data Analysis

“…With respect to program development, some generalizations have become possible with the improvement in the capability of equipment, increased price/performance of the equipment, and the availability of compiler level programming languages such as FORTRAN and ALGOL. In the area of statistical computing, the generalization of programs has been inhibited by the lack of effective general purpose algorithms, but much progress has been made in the past few years for analyses involving linear models, see for example Bock (8), Cooper (23), Hartley (40), Nelder (69) and (70), Oliver (72), Schlater and Hemmerle (75), Wilkinson (89), and Yates and Anderson (93). For example, viewing many seemingly distinct designs of experiments or analyses of variances as special cases of a generalized linear hypothesis can lead to useful insight.…”

A Multivariate Analysis Computer Program

“…For all types of analysis, the data can be presented to the MAP program in various forms and on various input media. The commonly imposed restriction to fixed format unit-by-unit input from cards would be intolerable in a program that, again because of its interactive nature, must be able to communicate with other programs such as the General Survey Program (Anderson, 1966) and the General Factorial Program (Yates and Anderson, 1966).…”