Extrinsic Sources of Bias

In Chapter 7, it was shown that if the correlation exponent (q) exists, then the probability function F

(y) could be decomposed as (y)y(q), where (y) denotes the non-powerlaw part. The non-powerlaw behaviour induced by the function (y) was referred to as an intrinsic form of bias. In Chapter 8, methods to estimate

for q = 2; 3; , were investigated. Because of experimental limitations or deficiencies, the empirical probability

distribution of sampled interpoint distances can have quite different characteristics to that of F

(y) given by Equation 7.6. We are not referring to those sources of bias that are related to the particular estimator used as in Chapter 8. The extrinsic sources of bias referred to here are analogous to non-sampling error in sample survey methodology, are an inherent part of the data, but not characteristic of the underlying process (i.e., not intrinsic bias).