The name is absent

In general, only a small proportion of cases had missing data (< 5 %) even for the measures of
social background, which is as a result of the procedures for tracking children and good relations
with primary schools, as well as regular data quality checks of the EPPE 3-11 data management
team. Higher proportions of missing values occur for income-related variables like salary, socio-
economic status (SES) or the eligibility for free school meals (FSM), which is also an additional
low income indicator. A higher proportion of missing values for these kind of measures is a
typical response pattern also found in other survey studies. ¹⁰

Cognitive assessments

To take account of development and age, the study uses different assessment instruments for
cognitive outcomes at different time points:

• Year 1: NFER-Nelson Primary Reading Level 1 and Mathematics 6 tests

• Year 5: NFER-Nelson Primary Reading Level 2 and Mathematics 10 tests.

To ensure comparability over time, an internal standardisation and normalisation procedure was
applied. This procedure takes account of age effects within one school year. The scores
presented in this paper are internally standardised to a mean of 100 and a standard deviation of
15. Therefore all children scoring better than 100 at a certain time point are scoring at or above
the attainment level expected for their chronological age (belong to the upper half of the sample
of that assessment, controlling for age-effects). Due to the use of internally standardised
attainment scores, the scores can only be used to investigate the progress or improvement of
certain groups of children relative to the total EPPE 3-11 sample, but cannot be used to show
absolute progress over time.

Associations between Children’s Attainments in different Outcomes and over time
Correlations can be used to explore associations between children’s attainments in different
outcomes and over time.¹¹ Children’s attainments in the Year 5 assessments were positively
correlated (r=0.68), indicating those who do well in Reading generally also do well in
Mathematics at the end of Year 5. The correlation between Reading and Mathematics scores at
the end of Year 1 (standardised test scores) was somewhat weaker (r = 0.58, not shown in Table
A1.2).

Table A1.2: Correlations between children’s standardised cognitive outcomes and with prior
assessments

The cognitive attainments are not only highly associated with each other but also show moderate
to high correlations with prior attainments (see Table A1.2). A particularly strong relationship is
found for attainment in Mathematics in Year 1 and Year 5 (r = 0.65), but also attainment in
Reading in Year 5 is fairly highly correlated with Reading attainment in Year 1 (r = 0.57). At this

¹⁰ To prevent loss of sample size for further analyses missing values for number of siblings, FSM and SES
where imputed using ‘the last observation carried forward’ method. Please see Sammons et al., 2007 for a
description of this imputation method. Family SES was calculated by combining mother’s and father’s
occupational categories and recording the higher of the two (family SES data was missing for 1.3% of the
sample after imputation of missing values).

¹¹ A correlation is a measure of statistical association that ranges form + 1 to -1.

42

<<   <   50   51   52   53   54   55   56   >   >>