1.5.2

Informal inference with percentages

Percentages, such as completion rates, need to be compared on the basis of their standard errors. This is similar to a test of a difference between two means, which uses the standard errors of the means. It is possible to make a rough comparison of two percentages by using the "average" confidence interval associated with them. In particular, if two percentages differ by more than a multiple of their average confidence interval then there is some evidence that they may be different. As explained in the Statistical Appendix (§6.2), this approximation is not valid when the sample sizes are small and other methods should be used.

1.5.3

Shifting denominator problem

As in the validation section of the Data Processing Appendix (§6.1), there were minor problems experienced with the data files across surveys in terms of linkage of student's records, and so there are minor inconsistencies in the totals between some sections of this report. Also, as students did not always fill in the answers expected of them, and in some cases filled in sections that they were logically not meant to, the denominator shifts between items. This is particularly the case with the employment section of the Point of Entry survey.