The Measurement of Irregularity in Growth and Cropping

Various methods are considered for the measurement of irregular and biennial phenomena.

An electronic computer was used to simulate bodies of data with known (i) basic pattern, whether biennial or not, (ii) trend, i.e. tendency to increase or decrease at a steady rate, (iii) random variation. Data were simulated:

(a) to test the various methods in their ability to detect a strict biennial effect in the presence of trend and random variation,

(b) to determine the statistical distribution of various methods of measuring bienniality and their reaction to patterns other than strictly biennial ones, and

(c) to ascertain the effect of length of period on the measures and their sensitivities.

The conclusion was reached that I, the established method of measuring bienniality, and a related quantity, K, were the most useful. By way of illustration, both were used with data from an experiment with apple trees. Until more is known, there is no reason to abandon I, though K has some attractive properties.