4.1.6. Processing errors

These errors affect the data set. They can arise by errors caused by the person who records the data (mistyping, copy/paste errors, stretching cells in an Excel spread sheet etc.). If different people capture the data, harmonisation of notation and careful procedures for recording data should be in place, checks should be made, and personnel should be well-trained and informed to avoid introducing errors and biases (Schaeffer et al., 1990).

Unlike the sampling error, the non-sampling errors are very difficult, if not impossible, to measure, and cannot be reduced by increasing the sample size. In a large sample, non-sampling errors are the more important source of errors, as the sampling error is reduced. The only way of controlling non-sampling error is to know what sorts of errors are possible, and to be very careful to avoid these as much as possible in the conduct of the survey.