Applied Multivariate Analysis of Student Academic Performance

his project applied multivariate techniques to the Students Performance in Exams dataset (1,000 students, 8 variables, no missing values). We:

summarised the variables with frequency tables, descriptive statistics, and graphs;

showed that the three score variables are strongly correlated and suitable for MANOVA;

ran MANOVA against five grouping variables and found highly significant multivariate effects for all of them;

followed up with univariate ANOVA, finding that the largest effects come from lunch type (socio-economic proxy), gender (especially in reading/writing), and test preparation (especially in writing).

Limitations. The dataset is observational, so the differences are associations, not causal effects. We do not have continuous covariates such as study hours or family income, so MANCOVA was not pursued. The lunch variable is only an indirect socio-economic proxy.

Future work. Add covariates (study time, prior GPA) and run MANCOVA. Apply PCA to compress the three scores into a single performance index. Test for interactions between gender and test preparation.