Growing up in Scotland: changes in child cognitive ability in the pre-school years

Footnotes

1. Although Cullis and Hansen (2008) consider other factors which influence cognitive ability at age 5, including maternal education, and also make use of value added models (which take account of earlier ability at age 3) in their analysis.

2.A description of the analysis is included in the Technical Appendix.

3.Further information on the design, development and future of the project is available from the study website: www.growingupinscotland.org.uk

4. The GUS sample is generated in two stages. The first stage randomly selects geographic areas or clusters, the second stage selects individuals within those clusters. The standard errors are adjusted to take account of the geographic clustering of the sample at the first stage.

5. The standard deviation provides a summary of the spread of data values around the mean.

6. This is measured as the highest educational qualification achieved by any of the child's parents who are resident in the household.

7. Measured using the National Statistics Socio-Economic Classification ( NS- SEC) and taken at household level - that is, the highest classification amongst all parents in the household. More information on NS- SEC is included in Appendix 1.

8. The method used by MCS researchers uses differences at the median, averaged across a range of ages. The resulting 'differences in age equivalents' are therefore very rough estimates and should be treated with caution.

9.That is, change in ability at the level of an individual child rather than change at an overall group level.

10. The statistical analysis and approach used in this report represents one of many available techniques capable of exploring this data. Other analytical approaches may produce different results from those reported here. A description of the analysis is included in the technical appendix.

11. The regression output is included in Tables 1 and 2 in the Technical Appendix.

12. A description of the analysis is included in the Technical Appendix.

13. Around one-third of the GUS children had started primary school by the time of their assessment at 58 months.

14. Full results from the regression are shown in Tables 3 and 4 in the Technical Appendix.

15. Our measure of family type takes into account changes in family type over time. It incorporates data on parents/carers in the household from the first three waves of data collection to identify, for example, where a couple family have separated or remained, or a lone parent becomes re-partnered.

16. Full results of the regression analysis are shown in Tables 5 and 6 in the Technical Appendix.

17. Home learning activities include, for example, reading with the child, painting and drawing, and singing nursery rhymes - a full description of the measure as used in this analysis is included in Appendix 1.

18. Full regression results are shown in Tables 7 and 8 in the Technical Appendix.

19. The full results of the regression are shown in Tables 9 and 10 in the Technical Appendix.

20. Due to the spread of birth dates amongst the sample and the timing of fieldwork, around one-third of children had started P1 by the time of the 58 month interview.

21. Difficulties with early language and communicative development was assessed at 22 months using the infant/toddler checklist of the Communication and Symbolic Behaviour scales. A lower score suggests less advanced development. Further details are included in Appendix 1.

22. Full regression results are shown in Tables 11 and 12 in the Technical Appendix.

23. More detail on differences in parental service use and support amongst families with young children is provided in
an accompanying GUS report: Mabelis, J. and Marryat, L. (2011) Growing Up in Scotland: Parental service use and informal networks in the early years.

24. Full regression results are shown in Tables 13 and 14 in Appendix 2.

25. The results of the full models are shown in Tables 15 and 16 in the Technical Appendix.

26. The full results are shown in Tables 19 and 20 in the Technical Appendix.

27. This issue is known as 'multicollinearity'. It occurs when two or more of the independent variables in a regression model are highly correlated. Multicollinearity can mean that the correlated variables may not be reporting unique or independent values. This can cause potentially misleading results for example indicating that an association is significant when it is not (because it is reporting the association with the correlated variable).

28. Full results of the regression are shown in Tables 21 and 22 in the technical appendix.

29. Combining the groups was necessary to maintain a large enough number of cases for consideration.

30. Full results are shown in Tables 23 and 24 in the Technical Appendix.

31. 'Reliability' is used here to denote the internal consistency of items making up a parenting measure. Consistency is estimated using Cronbach's alpha, which is based on the average correlation between items. The value of Cronbach's alpha depends in part on the number of items in the scale, with a greater number of items resulting in higher alphas. While there is no firm consensus, a commonly accepted 'cut-off' of an alpha of 0.7 or more for items to be included in a scale is often lowered to 0.6, particularly for exploratory studies.

32. Full regression results are shown in Tables 7 and 8 in the Technical Appendix.