Page 34 - Šolsko polje, XXVII, 2016, no. 3-4: IEA ICILS in druge sodobne teme, ur. Eva Klemenčič
P. 34
šolsko polje, letnik xxvii, številka 3–4
Educational Systems Indi- (SE) School (SE)
vidual SES
Thailand SES
Turkey
International Average 4.08 * (0.32) 7.56 * (0.59)
3.21 * (0.31) 5.96 * (0.73)
2.50 * (0.06) 4.86 * (0.14)
*Significant at p<.05 level
†Not meeting the sampling requirements
As the table shows, statistically significant CIL gaps between stu
dents in regard to their individual SES is found in all 20 countries in this
study. The gap is lowest in Hong Kong and highest in Thailand, 1.20 and
4.08 score points per unit increase in SES respectively. This is similar for
the school-level aggregated SES where in all countries the effect is positive
and significant, from 2.14 (Slovenia) to 7.56 score points (Thailand).
The control variables were added to the model to test the mitigation
effect of the individual and school characteristics (see the Method sec
tion). The results show that the individual and school-level SES gaps are
not mitigated when any of the variables in the Appendix are controlled
for, the effect of SES remains strong and significant (the results are not
published here due to the vast amount of outputs). These results (strong
relationship between individual SES and CIL and even stronger between
CIL and school SES) suggest that there are compositional and interaction
effects due to the individual and school-level SES. Additional Hierarchi
cal Linear Models were built to test these effects.
First, a null model is built to obtain the Intra-Class Correlation Co
efficients (ICCs). The model is presented in the equation below.
(L1) (1)
(L2),
where
– 1st to 5th PV of CIL
– intercept, expected achievement of student i within cluster j
– error at individual level
– CIL grand mean
– deviation of clusters around the grand mean
The ICCs of the CIL scores are presented in Table 3.
32
Educational Systems Indi- (SE) School (SE)
vidual SES
Thailand SES
Turkey
International Average 4.08 * (0.32) 7.56 * (0.59)
3.21 * (0.31) 5.96 * (0.73)
2.50 * (0.06) 4.86 * (0.14)
*Significant at p<.05 level
†Not meeting the sampling requirements
As the table shows, statistically significant CIL gaps between stu
dents in regard to their individual SES is found in all 20 countries in this
study. The gap is lowest in Hong Kong and highest in Thailand, 1.20 and
4.08 score points per unit increase in SES respectively. This is similar for
the school-level aggregated SES where in all countries the effect is positive
and significant, from 2.14 (Slovenia) to 7.56 score points (Thailand).
The control variables were added to the model to test the mitigation
effect of the individual and school characteristics (see the Method sec
tion). The results show that the individual and school-level SES gaps are
not mitigated when any of the variables in the Appendix are controlled
for, the effect of SES remains strong and significant (the results are not
published here due to the vast amount of outputs). These results (strong
relationship between individual SES and CIL and even stronger between
CIL and school SES) suggest that there are compositional and interaction
effects due to the individual and school-level SES. Additional Hierarchi
cal Linear Models were built to test these effects.
First, a null model is built to obtain the Intra-Class Correlation Co
efficients (ICCs). The model is presented in the equation below.
(L1) (1)
(L2),
where
– 1st to 5th PV of CIL
– intercept, expected achievement of student i within cluster j
– error at individual level
– CIL grand mean
– deviation of clusters around the grand mean
The ICCs of the CIL scores are presented in Table 3.
32
