Latent Class Logistic Analysis (Clustering Indonesian Students Achievement In Mathematics Base On TIMSS’ Survey).

Abstract

Conventional methods of clustering become weak when meet measured objects with qualitative or categorical data. Latent class logistic analysis can be an alternative method of clustering to overcome this problem. This research is aim to see the application of latent class logistic analysis to cluster the measured objects with qualitative and quantitative variable and at once to find out backgrounds of the clusters. The objects in this research are 2171 eight grade students from 133 schools in Indonesia. There are two results in this research; first in clustering and second in logistic analysis. In clustering, the students have been clustered into four ideal clusters, e.g. 39.16 percent students were in cluster1, 32.42 percent in cluster2, 21.46 percent in cluster3, and 6.97 percent in cluster4. Each cluster represents the students with very low, low, medium, and high ability in mathematics. In logistic analysis, overall, each cluster has been explained well by covariates e.g. student’s interest, attitude, aptitude and motivation on mathematics, parent’s social-economic condition, parent’s highest education level, teacher’s highest education level, teacher’s major study of mathematics and educations, teacher’s perceptions on schools, school’s facilities, etc.