This appendix is provided to give the reader more details of the comparisons made between the groups. It cannot be emphasized too strongly that the essential essence is that all comparisons of variable categories are made with respect to the baseline category for that variable (e.g. 0 hours worked per week).
The membership of a group (e.g. those for whom reading was a goal, as opposed to those for whom reading was not a goal) is reviewed with respect to a selection of the Point of Entry variables, First every variable is examined on an individual basis (univariately) (§4.1.10) to see if it differs between group members and non-members. Then using this knowledge, all the variables are jointly compared (§4.1.11) to obtain an artificial profile of the group members.
Where the Point of Entry variable has two values (e.g. gender), the results such as:
are straightforward to interpret - there are more females (62%: 56%) for whom reading is a goal. This is denoted by a bullet () in the data summaries. If the result is significant at the p <=0.01 level, then the bullet is also shaded.
However, when the Point of Entry variable can take more than two values the problem of determining where the differences lie is more complex. For example consider the goal of reading with respect to age, which is broken into 3 categories (15-19, 20-24, >=25 years of age).
With these multi-valued variables, a blanket or omnibus test is first done to see if there are any differences between the groups with respect to the categories as a whole. If there are differences (as there are here), further testing is warranted, otherwise no further comparisons are made, as this would amount to data dredging.
|Previous||Table of Contents||Next|