This thesis analyzes the Master careers of engineering students enrolled at Polytechnic of Milan, with particular focus on the international ones. The aim is to predict career outcomes whose classification is determined by Time-to-degree and GPA. Given the hierarchical nature of the dataset and the categorical variable of interest, multinomial multilevel regression models are adopted. Models are fitted under different settings, considering distinct nesting schemes and subsequent information in time. In particular, by exploring the clustered structures induced by both the degree programs and the countries in which students obtained their bachelor degree, an investigation of the heterogeneity at these levels is conducted. Moreover, multilevel models are fitted first by considering only personal data available at enrollment and, then, by adding the variable which synthesises the exam achievements obtained during the first semester of the Master career. Results highlight which are the students characteristics associated to their career output and the performance differentiation arising from the various nested structures which characterizes engineering Master graduates.
Questa tesi analizza le carriere magistrali degli studenti di ingegneria iscritti al Politecnico di Milano, con particolare attenzione a quelli internazionali. L’obiettivo è quello di prevedere gli esiti delle carriere, la cui classificazione è determinata dal tempo impiegato per il conseguimento della laurea e dalla media dei voti. Data la natura gerarchica del dataset e data la variabile categorica di interesse, sono stati utilizzati modelli di regressione multinomiale multilivello. Vengono prodotti più modelli, i quali sono basati su vari schemi di annidamento e su differenti variabili relative agli studenti. In modo particolare, l’eterogeneità intrinseca nelle osservazioni viene analizzata mettendo in risalto i raggruppamenti individuati dai corsi di laurea magistrale e dai paesi in cui gli studenti hanno conseguito la laurea triennale. I modelli multilivello, inoltre, sono stati prodotti considerando, prima solo i dati personali disponibili al momento dell’iscrizione e, successivamente, aggiungendo anche le informazioni riguardanti gli esami sostenuti durante il primo semestre della carriera magistrale. I risultati evidenziano quali sono le caratteristiche degli studenti associate agli esiti delle loro carriere e la differenziazione di performance scaturita dalle varie strutture gerarchiche che caratterizzano i laureati magistrali in ingegneria.
Multinomial multilevel models for predicting Master students careers at Politecnico di Milano
Tranchida, Alessio
2021/2022
Abstract
This thesis analyzes the Master careers of engineering students enrolled at Polytechnic of Milan, with particular focus on the international ones. The aim is to predict career outcomes whose classification is determined by Time-to-degree and GPA. Given the hierarchical nature of the dataset and the categorical variable of interest, multinomial multilevel regression models are adopted. Models are fitted under different settings, considering distinct nesting schemes and subsequent information in time. In particular, by exploring the clustered structures induced by both the degree programs and the countries in which students obtained their bachelor degree, an investigation of the heterogeneity at these levels is conducted. Moreover, multilevel models are fitted first by considering only personal data available at enrollment and, then, by adding the variable which synthesises the exam achievements obtained during the first semester of the Master career. Results highlight which are the students characteristics associated to their career output and the performance differentiation arising from the various nested structures which characterizes engineering Master graduates.File | Dimensione | Formato | |
---|---|---|---|
2023_05_Tranchida.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: testo tesi
Dimensione
2.85 MB
Formato
Adobe PDF
|
2.85 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/212117