Options
Dimensionality of school compositions and performance of Singapore secondary school students: a multivariate approach
Author
Tan, Teck Kiang
Supervisor
Kang, Trivina
Abstract
School composition is a crucial factor in understanding student performance. However, it is often treated as a unidimensional construct. This dissertation shows that a group of similar school compositions can be classified and described using multivariate models to depict the multidimensionality of school compositions. Four dimensions of school composition are identified from a list of 22 school compositions from a national representative sample of Singapore secondary school students. The findings on the dimensionality of school composition support the classification scheme developed to classify school compositions into two main types of school composition: representativeness and diversity. The former represents the average students in a school and the latter represents the spread of students’ social composition.
The results of the multilevel models show that different representative and diversity school compositions are associated with students’ performance at different levels of effect sizes. In particular, the representativeness of SES school composition explained up to 94% of the school variance, indicating that variance between schools’ performance are closely associated with their students’ family background. Stream composition, diversity in socio-psychological composition, and language composition are also influential school compositions that explain at least 60% of school variances in student academic performance in English.
A new methodological approach to organize the presentation of empirical findings in a systematic way named as “blocks of interest approach” is introduced in this dissertation. This analytic strategy shows that by extracting findings in a systematic way helps variable selection during modeling, and in comparing and contrasting results from different models so as to conduct scenario analysis.
It is observed in the analysis that student level variables are able to explain substantial school level variances. This dissertation names this phenomenon “duality property of variables” and it is partially explained by what I have termed the “historical content” of the variable. The historical content of a variable refers to the historical information that is contained within the variable that conveys the implicit conceptual meaning of the variable.
The results of the multilevel models show that different representative and diversity school compositions are associated with students’ performance at different levels of effect sizes. In particular, the representativeness of SES school composition explained up to 94% of the school variance, indicating that variance between schools’ performance are closely associated with their students’ family background. Stream composition, diversity in socio-psychological composition, and language composition are also influential school compositions that explain at least 60% of school variances in student academic performance in English.
A new methodological approach to organize the presentation of empirical findings in a systematic way named as “blocks of interest approach” is introduced in this dissertation. This analytic strategy shows that by extracting findings in a systematic way helps variable selection during modeling, and in comparing and contrasting results from different models so as to conduct scenario analysis.
It is observed in the analysis that student level variables are able to explain substantial school level variances. This dissertation names this phenomenon “duality property of variables” and it is partially explained by what I have termed the “historical content” of the variable. The historical content of a variable refers to the historical information that is contained within the variable that conveys the implicit conceptual meaning of the variable.
Date Issued
2014
Call Number
LA1239.7 Tan
Date Submitted
2014